A Biomedical Researcher's Guide for Analyzing Short Tandem Repeat (STR) Genotypes of Human Cell Lines and in Vitro Tissue Samples Using Three Standard Authentication Algorithms

Cell line authenticity is critical for producing reproducible and rigorously valid biomedical research. As of version 14 of its Register of Misidentified Cell Lines, released in 2026, the International Cell Line Authentication Committee (ICLAC) has identified 560 commonly used cell lines that are misidentified and for which there are no known authentic samples, and 48 samples of misidentified cell lines for which authentic samples have been found. The Cellosaurus Cell Line Knowledge Resource of August 2025 lists information about 122,819 human‐derived cell lines, of which 1396 are “problematic cell lines.” From extensive surveys of frequently used cell lines around the world, on average 22% of cell lines, or 2 of every 9, being used in laboratories are likely to be incorrect. Consequently, it is important that cell lines be authenticated (1) before starting a project utilizing them, (2) during their use in a project, and (3) after the project is completed to avoid publishing invalid and irreproducible information. The authentication of human cell lines is best achieved through analysis of short tandem repeat (STR) sequences in the genomic DNA. The resulting genotype data must be compared with those of reference genotypes in databases to determine whether they are derived from the original tissue sample or are misidentified. Cellosaurus release 54.0 contains 9063 STR genotypes for human‐derived cell lines. This report with its Supporting Information, along with the accompanying workflow protocol article, are designed to provide an explanation of how to acquire and best apply three standard algorithms for correctly analyzing STR authentication genotyping data of human cell lines and cultured tissue samples. Unlike the revised Human Cell Line Authentication Standardization of STR Profiling (ASN‐0002), this article is aimed at researchers, editors, and reviewers who need to understand how to interpret the resulting comparisons. This article is related to the accompanying protocol article entitled “Analytical Workflow Protocol for Determining the Authenticity of Human Cell Lines and Tissue Samples by Genotype STR Analysis” (Korch et al., submitted). © 2026 Wiley Periodicals LLC.

authentication algorithms, cell line and tissue authentication, human cell lines, STR genotyping

Cell lines and xenografts are extensively used as models for biological processes in multiple disciplines, including biomedical research. For these to serve as valid and meaningful models, it is critical that they be correctly identified as coming from not only the correct species but also from the correct donor sample. Mammalian cell lines, such as those derived from human organs or different types of cancer, often grow slowly. If cultures of such cell lines are cross‐contaminated with faster‐growing cell types, they can be taken over and thus no longer represent the original model. Genetic drift due to different mechanisms, such as errors in DNA mismatch repair (MMR), can also give rise to genetic variants of cell lines. Even cell lines with seemingly identical short tandem repeat (STR) genotypes (also known as STR profiles or STR fingerprints) can, within as little as three passages, genetically drift from each other sufficiently to show significant phenotypic differences (Ben‐David et al., 2018; Kleensang et al., 2016). Technical mistakes, such as sample labeling errors, and failure to confirm cell line and tissue identities throughout projects have often resulted in research being done with the wrong samples (Huang et al., 2017; Korch et al., 2012, 2018; Masters et al., 2001; Schweppe et al., 2008; van Bokhoven et al., 2001, 2003; Vaughan et al., 2017; Ye et al., 2015; Yu et al., 2015; Zhao et al., 2011). An example of the consequences for the scientific literature is the use of the false cell line KB (CVCL_0372), which was shown in the 1960s to actually be HeLa cells, which were derived from a cervical adenocarcinoma. Ninety percent of the publications between 2000 and 2014 using this cell line described it with its false identity of oral epidermoid carcinoma (Vaughan et al., 2017) and used it to model oral carcinomas. The research articles that cited these original articles, which were based on the false identity of KB, tended to describe this line similarly. There was no significant difference between the impact factors of the articles describing KB incorrectly versus those describing it correctly; however, the validity of the former research is clearly questionable. Figure 1 summarizes the history of a cell line that has been used under three false identities and contributed to three fields of cancer biology—breast cancer, thyroid cancer, and uveal melanoma. The M14 cell line was established in 1975 from a metastatic melanoma on the buttocks of a man. It was shared with several laboratories and clearly took over multiple cell lines listed in Figure 1. In the early 2000s, it was classified as being of melanoma origin by gene expression analyses, but various groups did not perform STR analysis of their various imposter cell lines until 2007‐2008, when several supposed thyroid cell lines were shown to be actually M14 (Schweppe et al., 2008). The debate about the identity of the imposters listed in Figure 1 continued until early samples of M14 dating from 1975—that is, before the establishment of the other lines—were analyzed for STRs and found to be derived from a male and to have STR genotypes matching those listed in this figure. The M14 karyotype in Figure 1 shows that probing for Yq12 chromosomal DNA indicates that the cell line had lost part of the Y chromosome, but retained a portion of the Yq12 region through its translocation to the autosomal chromosome 22, whereby two copies of Yq12 DNA were retained as a remnant of the Y chromosome from the original tissue sample (Korch et al., 2018). The importance of authenticating cell samples is reviewed more extensively in the accompanying protocol article (Korch et al., submitted). The current “gold standard” for checking cell line identity by STR genotyping was first demonstrated in 1999 (Tanabe et al., 1999). This technique, originally developed for forensic purposes, determines which allelic variants at multiple STR loci consisting primarily of either tetra‐ or pentanucleotide repeats are borne by samples of human‐derived cell lines and tissues. In biomedical research, it is important to be able to confirm, for example, that any behavioral changes observed in a cell line culture are due to phenotypic responses of the actual original culture and not to the behavior of unrelated cells that have cross‐contaminated and taken over the test culture. Therefore, confirmation of the genetic identity of cell lines is essential to the validity and reproducibility of experimental results. The accompanying protocol article (Korch et al., submitted) and other resources describe the generation of STR data for cell samples and how to access their quality. STR genotyping technology, as applied to cell lines, is described in the accompanying protocol article and in greater detail elsewhere (Korch et al., submitted; Almeida & Korch, 2023; Capes‐Davis & Freshney, 2021; Korch et al., 2022). In brief, the current STR genotyping technology is based on a multiplex PCR amplification of 8‐26 STR loci in the human genome using primers tagged with one of up to six different fluorescent dyes. The resulting amplicons are separated by capillary electrophoresis, and the number of repeats in the different loci is deduced from the sizes of the amplicons for the different STR loci. In 2022, the American National Standards Institute (ANSI) and the American Type Culture Collection (ATCC) published the second edition of the ASN‐0002 standard methodology for human cell line authentication with detailed instructions and extensive troubleshooting guidelines (Korch et al., 2022). Also in that year, the U.S. National Center for Biotechnology Information released a comprehensive update of an Assay Guidance Manual describing how to authenticate both human and mouse cell lines and tissues (Almeida & Korch, 2023). As mentioned, STR genotyping deduces the number of repeats in STR alleles at many loci through a multiplex PCR assay, with determination of the STR alleles by analysis of amplicon size at multiple STR loci. Several mathematical algorithms for comparing STR profiling data and identifying related cell lines have been described in the literature (Bady et al., 2012; Capes‐Davis et al., 2013; Eltonsy et al., 2012; Lorenzi et al., 2009; Masters et al., 2001; Romano et al., 2009; Romano et al., 2016; Somaschini et al., 2013; Tanabe et al., 1999). These algorithms are described and compared in Appendix E of the ANSI‐ATCC ASN‐0002 standard published (Korch et al., 2022). The various recommendations in the ASN‐0002 2022 standard for authenticating cell lines by STR genotyping and the analysis of the resulting profiling data are supported by 25 years of publications in this area (Bady et al., 2012; Capes‐Davis et al., 2013; Huang et al., 2017; Korch & Capes‐Davis, 2021; Korch et al., 2012, 2018; Korch & Varella‐Garcia, 2018; Masters et al., 2001; Schweppe et al., 2008; Tanabe et al., 1999; van Bokhoven et al., 2001, 2003; Yu et al., 2015; Zhao et al., 2011). Ideally, an algorithm used for comparing STR profiles and determining whether two cell lines or tissue samples are derived from the same original donor should meet the following eight goals. Goal 1: It can determine the true identity of misidentified cell lines based on all available data; Goal 2: It can be easily understood and applied by “lay” researchers, and the results can be easily explained to researchers by the STR profiling technical staff; Goal 3: It can produce identical results when comparing cell line A to cell line B and conversely B to A; Goal 4: It can identify at least one of the cell lines in a test sample that is a mixture of two or more lines; Goal 5: It can identify the original cell line from which a subline has evolved due to genetic drift in a culture of cells caused by defects in DNA replication, for example, microsatellite instability (MSI); Goal 6: It does not count the amelogenin alleles because they are not STR alleles and the Y‐linked amelogenin allele (AmeloY), which is used for determining the gender of the donor, is often lost during the establishment and culturing of male‐derived cell lines (Yu et al., 2015); Goal 7: It does not assume that single STR peak alleles represent homozygosity. Such peaks should be counted only as single alleles and not twice, as is done for forensic samples of DNA from normal diploid cells. This is because cultured human cell lines often undergo loss of heterozygosity (Yu et al., 2015), and an STR genotyping assay cannot verify the number of homozygous alleles in a culture. Also, cultures of the same cell line may contain subpopulations of cells with different levels of ploidy; and Goal 8: It does not use peak height (i.e., signal strength) to estimate ploidy because the PCR reactions and capillary electrophoretic conditions are not quantitative. This report demonstrates the best uses of three simple STR analysis algorithms that, when used together, can meet these eight goals. This protocol is targeted to biomedical researchers to provide an understanding of how genetic drift, DNA MMR defects, degree of heterozygosity, cross‐contamination, and mishaps in labeling, handling, and culture contribute to the spectrum of human cell line and tissue variability and the potential limitations on their utility for research projects. It is also targeted to the editors of biomedical journals, the gatekeepers of such research. It is essential that journals implement more stringent training for their editors and reviewers on the critical evaluation of cell line authenticity data. As of January 2026, the most comprehensive policy is that of the International Journal of Cancer. METHODOLOGY The simplest and most frequently used and described STR match algorithm in the literature, which is also the most easily applied by researchers and service centers, is the Tanabe et al. (1999) modification of the Dice (1945) and Sørensen (1948) equations. These were originally applied to plant populations. The original Sørensen quotient of similarity (QS) equation was defined as: This formula states that the QS is equal to the number of shared characteristics (C) divided by the average number of characteristics in shared properties in the two population samples (A and B). In 1999, this formula was rearranged by multiplying the numerator and denominator by 2 (Tanabe et al., 1999) and was applied to STR profiles of cell lines to identify related cell lines, using data only at shared loci, by the following formula: In 2001, Masters et al. (2001) applied a different algorithm to estimate the similarity between two STR profiles. It calculated the number of shared alleles as a percentage of the alleles in the query profile at shared loci: In 2013, Capes‐Davis et al. (2013) compared the similarity estimates resulting from using the above two algorithms with those obtained using an alternative to the Masters algorithm—namely, where the denominator is not the number of alleles in the query profile, but rather the number of alleles in the reference profile at shared loci, that is: This article identifies when and how to apply these three formulae for comparing STR genotype profiles of human cell lines using the CLASTR search tool available on the Cellosaurus website (www.cellosaurus.org/str‐search/) and the ICLAC STR profile comparison tool found in Appendix 2 of the accompanying protocol article (Korch et al., submitted).

OBTAINED USING THREE DIFFERENT MATCH ALGORITHMS Tanabe‐Sørensen (TNB) Algorithm HeLa cell lines The Tanabe (TNB) algorithm works very well in most cases: namely, when (1) the numbers of alleles in the two sample profiles are similar, (2) the cell culture is not a mixture of cell lines, (3) the cell line is MSI stable, and (4) the genotypes have not drifted greatly genetically. As an example, Table 1 illustrates the similarity of the genotype of the cervical adenocarcinoma HeLa cell line genotype with those of eight well‐known imposter cell lines cross‐contaminated and taken over by HeLa cells. These imposters were identified by Stanley Gartler in 1967‐1968 (Gartler, 1967, 1968), and yet they continue to be used under their false identities (Horbach & Halffman, 2017; Korch & Capes‐Davis, 2021; Vaughan et al., 2017). Table 1. | Cell line | Match score | AMEL | CSF1PO | D2S1338 | D3S1358 | D5S818 | D7S820 | D8S1179 | D13S317 | D16S539 | D18S51 | D19S433 | D21S11 | FGA | Penta D | Penta E | TH01 | TPOX | vWA | |---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---| | HeLa CVCL_0030 | Reference with HeLa | X | 9,10 | 17 | 15,18 | 11,12 | 8,12 | 12,13 | 12,13.3 | 9,10 | 16 | 13,14 | 27,28 | 18,21 | 8,15 | 7,17 | 7 | 8,12 | 16,18 | | HeLa CVCL_0030 | TNB = 100% MvQ = 100% MvR = 100% | X | 9,10 | 17 | 15,18 | 11,12 | 8,12 | 12,13 | 12,13.3 | 9,10 | 16 | 13,14 | 27,28 | 18,21 | 8,15 | 7,17 | 7 | 8,12 | 16,18 | | HEp‐2 CVCL_1906 | TNB = 100% MvQ = 100% MvR = 100% | X | 9,10 | 17 | 15,18 | 11,12 | 8,12 | 12,13 | 12,13.3 | 9,10 | 16 | 13,14 | 27,28 | 18,21 | 8,15 | 7,17 | 7 | 8,12 | 16,18 | | Intestine 407 CVCL_1907 | TNB = 100% MvQ = 100% MvR = 100% | X | 9,10 | 11,12 | 8,12 | 12,13 | 12,13.3 | 9,10 | 16 | 27,28 | 18,21 | 7 | 8,12 | 16,18 | ||||| | Girardi Heart clone 7 CVCL_1816 | TNB = 100% MvQ = 100% MvR = 100% | X | 9,10 | 17 | 15,18 | 11,12 | 8,12 | 12,13 | 12,13.3 | 9,10 | 16 | 13,14 | 27,28 | 18,21 | 8,15 | 7,17 | 7 | 8,12 | 16,18 | | FL CVCL_1905 | TNB = 100% MvQ = 100% MvR = 100% | X | 9,10 | 11,12 | 8,12 | 12,13.3 | 9,10 | 7 | 8,12 | 16,18 | ||||||||| | KB CVCL_0372 | TNB = 98.4% MvQ = 96.8% MvR = 100% | X | 9,10 | 17 | 15,18 | 11,12 | 8,12 | 12,13 | 12,13.3 | 9,10 | 16 | 13,14 | 27,28 | 21 | 8,15 | 7,17 | 7 | 8,12 | 16,18 | | KB‐3‐1 CVCL_2088 | TNB = 98.4% MvQ = 96.8% MvR = 100% | X | 9,10 | 17 | 15,18 | 11,12 | 8,12 | 12,13 | 12,13.3 | 9,10 | 16 | 13,14 | 27,28 | 21 | 8,15 | 7,17 | 7 | 8,12 | 16,18 | | Chang Liver CVCL_0238 | TNB = 98.2% MvQ = 96.4% MvR = 100% | X | 9,10 | 15,18 | 11,12 | 8,12 | 12,13 | 12,13.3 | 9,10 | 16 | 27,28 | 21 | 8,15 | 7,17 | 7 | 8,12 | 16,18 | || | WISH CVCL_1909 | TNB = 96.2% MvQ = 92.6% MvR = 100% | X | 9,10 | 17 | 15,18 | 11,12 | 8,12 | 12,13 | 13.3 | 9,10 | 16 | 13,14 | 27,28 | 21 | 7 | 8,12 | 16,18 | || | Selected hybrid cell lines | ||||||||||||||||||| | CNE‐1 CVCL_6888 | TNB = 72.4% MvQ = 67.7% MvR = 80.8% | X | 10,11 | 17,23 | 15,18 | 11,12 | 10,12 | 12,16 | 10,12, 13.3 | 9,10 | 13,16 | 13 | 27,30 | 18,21 | 6,7,9 | 8,12 | 14,16 | || | HNE‐1 CVCL_0308 | TNB = 66.7% MvQ = 61.0% MvR = 75.9% | X | 10,11 | 17,23 | 15,18 | 11,12 | 10,12 | 12,17 | 10,12,13.3 | 9,10,11 | 13,16 | 13 | 27,30 | 18,21 | 9,12 | 17,20 | 6,7,9 | 8,12 | 14,16 | | Sacc‐83 CVCL_h589 | TNB = 63.3% MvQ < 60% MvR = 75.0% | X | 10,11 | 17,23 | 15,16,18 | 11,12 | 10,12 | 12,16,17 | 10,12, 13.3 | 9,10 | 13,16 | 13 | 27,30 | 18,21 | 9,12 | 14,16 | ||| | SUNE2 CVCL_6956 | TNB = 63.0% MvQ < 60% MvR = 73.1% | X | 10,11 | 17,23 | 15,18 | 11,12,13 | 10,12 | 12,13,17 | 10,12, 13.3 | 9,10 | 13,16 | 13 | 27,30 | 18,21 | 9,12 | 17,20 | 6,7,9 | 8,9 | 14,16 | The comprehensive set of HeLa STR genotypes in Table S1 is the source of the selected data in this table. The genotypes of the imposter cell lines were probably derived from cross‐contamination of the original cell lines that were then taken over by HeLa cells. The four HeLa fusion hybrid cell lines were most likely derived from fusion of HeLa cells with a different human cell line of unknown identity. Sublines have been derived from one of these hybrid cell lines. The values in the Match score column are the matches to the HeLa CVCL 0030 STR profile calculated using the three algorithms in the CLASTR tool (Tanabe [TNB], Masters versus Query [MvQ], and Masters versus Reference [MvR]). The highlighted alleles in D13S317 and FGA loci indicate loci that have lost an allele from some strains as compared with the HeLa genotype. All STR loci in the hybrid cell lines have two or more alleles at all but the D19S433 locus, supporting the hypothesis that there is more than one genome in these cell lines. This hypothesis is also supported by the fact that the MvR scores are the highest of those determined by the three algorithms. Many more profiles of HeLa cells and related imposter cell lines are listed in Supporting Information Table S1, ranked by their similarity as determined using the TNB scoring algorithm with the HeLa CVCL_0030 STR profile as the reference genotype. Using this algorithm, the general recommendation in the literature is that, depending on the number of loci tested, scores of >80% or >90% will identify related strains of cell lines derived from the same donor (Bady et al., 2012; Capes‐Davis et al., 2013; Korch et al., 2022; Masters et al., 2001; Yu et al., 2015). The fewer loci tested, the higher the cutoff level that should be used. As is clearly illustrated by these HeLa‐derived cell lines, related cell lines can be identified by this algorithm. Moreover, the STR genotypes of the eight imposter cells match the HeLa STR genotype by scores between 92.6% and 100% on the basis of the three different algorithms (TNB, MvR, and MvQ; Table 1), showing that all three algorithms will easily identify related cell lines when the STR profiles of the two cell lines being compared have similar number of alleles at the shared loci. The highlighted alleles in the D13S317 and FGA loci are the loci that in some strains have lost an allele. Listed at the bottom of Table 1 are four of the thirteen STR genotypes from Supporting Information Table S1 for cell lines that are most likely hybrids derived from fusions between HeLa cells and another, unidentified human cell line (Strong et al., 2014). All of the STR loci except D19S433 have two or more alleles, supporting the hypothesis that these cell lines include more than one genome. This hypothesis is supported, as explained below, by the fact that the MvR algorithm produced the highest scores of those determined by the three algorithms. A characteristic feature of HeLa‐derived cell lines is that nearly all samples have a 13.3 repeat allele at the D13S317 locus, which has sometimes been called as a 14‐repeat allele due to poorer resolution of older STR assay methods. This allele has been associated only with HeLa‐derived cell lines and only with two forensic samples. No instances of the allele have been found in 1817 forensic samples in the 2015 FBI Population Data for the expanded CODIS core STR loci (https://ucr.fbi.gov/lab/biometric‐analysis/codis/expanded‐fbi‐str‐2015‐final‐6‐16‐15.pdf). This allele has been found in one forensic sample reported by the Alabama Department of Forensic Sciences and another reported by the Forensic Genetics Department, Autonomous Police of the Basque Country of Spain, to the U.S. National Institute of Standards and Technology STRBase database (NIST STRBase v2.0; Butler et al., 2025). Furthermore, according to the curator of the STRider database of human STR alleles, the 13.3 allele at locus D13S317 has not been detected among 12,796 individual genotypes in release 4 of the STRidER database (https://strider.online/query), nor in other data not included in the online database. One common way to estimate a minimum allele frequency for rare or unknown alleles in forensic genetic calculations is 5/2n; thus, for the 13.3 allele at D13S317, this frequency estimate would be = 5/(2 × 12,796) < 0.0002 (Martin Bodner, Institute of Legal Medicine, Medical University of Innsbruck, Innsbruck, Austria, personal communication, September 29, 2025). DLD‐1 family of cell lines Another example of a family of related (i.e., synonymous) and misidentified cell lines are those derived from the cell line denoted in Cellosaurus by the identifier CVCL_3449—namely, the parent of DLD‐1 (CVCL_0248) and three of its misidentified derivatives: HCT‐8 (CVCL_2478), HCT‐15 (CVCL_0292), and HRT‐18 (CVCL_2514). The parent cell line was not genotyped, but the STR genotypes of these “child” lines were profiled both by an older method (Vermeulen et al., 1998) and by current methods. The newer STR results from Cellosaurus are presented in Table 2. A CLASTR search identified four additional related cell lines: C146 (CVCL_S958), C170 (CVCL_S959), MIP‐101 (CVCL_H689), and T‐47D‐Luc (CVCL_AS01). Earlier, T‐47D‐Luc had been shown to be misidentified and found to be actually derived from the DLD‐1 family of cell lines, but the correct identities of the other three lines had not previously been reported. Table 2. | Cell lines | DLD‐1 CVCL_0248 a | HCT‐8 CVCL_2478 | HCT‐15 CVCL_0292 b | HRT‐18 CVCL_2514 a | C146 CVCL_S958 | C170 CVCL_S959 | MIP‐101 CVCL_H689 | T‐47D‐Luc CVCL_AS01 | |---|---|---|---|---|---|---|---|---| | Cellosaurus donor gender, age, disease | Male, 67 yr, colon adenocarcinoma | Male, 67 years, colon adenocarcinoma | Male, 67 yr, colon adenocarcinoma | Male, 67 yr, colon adenocarcinoma | Colorectal adenocarcinoma | Male, colorectal adenocarcinoma | Male, 54 yr, colon adenocarcinoma | Male, 67 yr, colon carcinoma | | Cellosaurus comments on cell line provenance | MSI unstable. Parent: CVCL_3449 is Parent cell line of DLD‐1, HCT‐8, HCT‐15, HRT‐18. Cellosaurus includes multiple STR profiles of related cell lines; only 4 are presented here. Children: DLD‐1 clone #1‐Luc (CVCL)J250). | MSI unstable. Parent: CVCL_3449 is Parent cell line of DLD‐1, HCT‐8, HCT‐15, HRT‐18. Cellosaurus includes multiple STR profiles of related cell lines (e.g., HCT‐8/VCR, HCT‐8‐Luc‐2tdT, HCT‐8‐tdT, HCT‐8‐mChe). | MSI unstable high. Parent: CVCL_3449 is Parent cell line of DLD‐1, HCT‐8, HCT‐15, HRT‐18. Cellosaurus includes multiple STR profiles of related cell lines; only 4 are presented here. Children: HCT15‐Luc#1 (CVCL_4Y02). | MSI unstable. Parent: CVCL_3449 is Parent cell line of DLD‐1, HCT‐8, HCT‐15, HRT‐18. Cellosaurus includes multiple STR profiles of related cell lines; only 4 are presented here. Children: HRT‐18G (CVCL_2515). | MSI unstable | MSI unstable | MSI unstable. Children: CVCL_C9CQ (MIP‐CEA clone 17), CVCL_C9CR (MIP‐CEA clone 6), CVCL_C9CS (MIP‐CEA clone 8), CVCL_C9CT (MIP‐neo), CVCL_HE61 (MIP/5FU), CVCL_HE62 (MIP/CIS), CVCL_HE63 (MIP/CPT), CVCL_HE64 (MIP/ETO). | MSI unstable. Parent: DLD‐1 family. Problematic cell line: Misidentified, originally thought to be from breast cancer cell line T‐47D, but has a Y chromosome and STR profile matches that of DLD‐1/HCT/HRT family of male‐derived colon cell lines. | | Genotype loci | |||||||| | Amelogenin | X,Y | X or X,Y | X,Y | X,Y | X | X,Y | X | X,Y | | CSF1PO | 11,12 or 12 | 12 | 12 | 11,12 or 12 | 12 | 12 | 12 | 12 | | D2S1338 | 17,25 | 17,25 | 17,25 | 17,25 | |||| | D3S1358 | 17 | 17 | 17 | 17 | 17 | ||| | D5S818 | 13 | 13 | 13 | 13 | 13 | 13 | 13 | 13 | | D7S820 | 10,12 c | 10,11.3 c or 10,12 c | 10,12 c | 10,11.3 c or 10,12 c | 10,12 c | 10,12 c | 10,11.3 c | 10,12 c | | D8S1179 | 15 | 15 | 15 | 15 | 15 | ||| | D12S391 | 19,22 | 19,22 | 19,22 | ||||| | D13S317 | 8,11 | 8,11 | 8,11 | 8,11 | 8,11 | 8,11 | 8,11 | 8,11 | | D16S539 | 12,13 | 12,13 | 12,13 | 12,13 | 12,13 | 12,13 | 12,13 | 12,13 | | D18S51 | 11,17 | 11,17 | 11,17 | 11,17 | 11,17 | ||| | D19S433 | 14,16 | 14,16 | 14,16 | 14,16 | |||| | D21S11 | 29,32.2 | 29,32.2 | 29,32.2 | 29,32.2 | 29,32.2 | 29,32.2 | 29,32.2 | | | FGA | 22 | 22 | 15,22 or 22 | 22 | 22 | ||| | Penta D | 9,14 | 9,14 | 9,14 | 9,14 | 9,14 | ||| | Penta E | 7,14 | 7,14 | 7,14 | 7,14 | 7,14 | ||| | TH01 | 7,9.3 | 7,9.3 | 7,9.3 | 7,9.3 | 7,9.3 | 7,9.3 | 7,9.3 | 7,9.3 | | TPOX | 8,11 | 8,11 | 8,11 | 8,11 | 8,11 | 8,11 | 8,11 | 8,11 | | vWA | 18,19 | 18,19 | 18,19 | 18,19 | 14,18,19 | 18,19 | 18,19 | 18,19 | | Tanabe score | Reference | 98.30% | 98.30% | 98.30% | 94.12% | 98.04% | 90.91% | 96.50% | The different STR genotypes were determined using different STR genotyping kits which did not query all of the same STR loci. This is reflected in the absence of data for some loci from some genotypes. The vWA locus of C146 cells has a 14 allele (red font), unlike any of the other samples. Does not include STR data for D6S1043 for DLD‐1 and HRT‐18, which are available in Cellosaurus. Does not include the STR data for HCT‐15 at D1S1656, D2S441, D10S1248, D22S1045, and DYS391, which are available in Cellosaurus. D7S820 alleles 11.3 and 12 differ by only 1 bp and may be the same allele depending on gel resolution used for the STR analysis. Some samples have the 11.3 allele. The STR genotypes in Table 2 illustrate several aspects of STR data. First, all lines have profiles that match that of DLD‐1 with TNB scores >90%. Second, even using STR genotyping kits testing as few as eight STR loci, one can still identify related cell lines. However, some older assay methods may not distinguish between some alleles that differ by a single base pair—for example, between the D7S820 alleles with 11.3 repeats and one with 12 repeats. The latter allele may actually be an 11.3 allele that is sometimes mistakenly called a 12 allele due to limited peak resolution during gel electrophoresis. This phenomenon is similar to that with the 13.3 allele of the D13S317 locus described above for HeLa cells. This is a commonly seen in data produced with older STR genotyping methods and can be resolved with newer kits, capillary electrophoresis, and more complete allelic ladders. Next, the MIP‐101 and C146 cell lines must have been derived from a male cell line of the DLD‐1/HCT/HRT family, but through culturing have lost the AmeloY allele. In addition, C146 has gained a 14‐repeat vWA allele. Finally, all child lines from the MIP‐101 line are expected to be misidentified, but because there are no STR data in Cellosaurus for any of the MIP‐1 child cell lines, this assumption cannot be confirmed. MCF‐7 cell lines Different cultures of a genetically stable cell line can show variations in their STR genotypes that vary during culturing. Supporting Information Table S2 contains 90 STR profiles for both synonymous cell lines identified as MCF‐7 (CVCL_0031) and for four sets of misidentified cell lines that have been taken over by MCF‐7. The sources of these genotypes are as follows: 34 are from this author's core facility (named CU‐AMC_DP0###‐##), 11 are from different published sources, 33 are listed in the Cellosaurus database, and 12 were found to be misidentified cell lines that had been taken over by MCF‐7. Table 3 shows nine examples of these profiles. The STR profiles were determined using different genotyping kits, and as a result not all loci were tested for all samples. Three alleles, at CSF1PO (allele 11), D3S1358 (allele 17), and D5S818 (allele 13), are very infrequent. In contrast, the FGA locus has three different alleles, 23, 24, and 25, which appear in various combinations in MCF‐7 samples. These different alleles at these four loci differ by one tetranucleotide repeat, suggesting that they could have arisen through DNA mismatch errors. Table 3. | Cell line | Match scores: | CSF1PO | D2S1338 | D3S1358 | D5S818 | D7S820 | D8S1179 | D13S317 | D16S539 | D18S51 | D19S433 | D21S11 | FGA | vWA | || |---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---| | TNB | MvR | MvQ | |||||||||||||| | MCF‐7 UC‐AMC DP0200‐04 |

| 10 | 21,23 | 16 | 11,12 | 8,9 | 10,14 | 11 | 11,12 | 14 | 13,14 | 30 | 23,24,25 | 14,15 | || | MCF‐7 CVCL_0031 | 100% | 100% | 100% | 10 | 21, 23 | 16 | 11,12 | 8,9 | 10,14 | 11 | 11,12 | 14 | 13,14 | 30 | 23,24,25 | 14,15 | | MCF7 CU‐AMC DP0197‐05 | 92.3% | 96.0% | 88.9% | 10 | 21, 23 | 16,17 | 11,12, 13 | 8,9 | 10,14 | 11 | 11,12 | 14 | 13,14 | 30 | 23,25 | 14,15 | | MCF‐7 CU‐AMC DP0286‐01 | 93.9% | 92.0% | 95.8% | 10,11 | 21, 23 | 16 | 12 | 8,9 | 10,14 | 11 | 11,12 | 14 | 13,14 | 30 | 23,25 | 14,15 | | MCF‐7 CU‐AMC DP0171‐05 | 95.8% | 92.0% | 100% | 10 | 21, 23 | 16 | 12 | 8,9 | 10,14 | 11 | 11,12 | 14 | 13,14 | 30 | 23,24 | 14,15 | | hTERT‐EEC‐A15‐1 CU‐AMC DP0135‐01 | 98.0% | 96.0% | 100% | 10 | 21, 23 | 16 | 11,12 | 8,9 | 10,14 | 11 | 11,12 | 14 | 13,14 | 30 | 23,25 | 14,15 | | ECC‐1 p146 CU‐AMC DP0128‐04A | 92.0% | 92.0% | 92.0% | 10 | 21,24 | 16 | 11,12 | 8,9 | 10,14 | 11 | 11,12 | 14 | 13,14 | 30 | 23,24,25 | 14,15 | | KPL‐1 CVCL_2094 | 98.0% | 96.0% | 100% | 10 | 21,23 | 16 | 11,12 | 8,9 | 10,14 | 11 | 11,12 | 14 | 13,14 | 30 | 23,25 | 14,15 | | BG1 (NIEHS) CU‐AMC DP0126‐04 | 100% | 100% | 100% | 10 | 21, 23 | 16 | 11,12 | 8,9 | 10,14 | 11 | 11,12 | 14 | 13,14 | 30 | 23,24,25 | 14,15 | | BG‐1 (from France) CVCL_6570 | 30.3% | 32.0% | 28.6% | 10,11 | 22,23 | 16,17 | 12 | 12 | 12,13 | 10 | 9,11 | 12,15 | 13,15 | 29,33.2. 34.2 | 20,24 | 14,17 | Included is a commonly circulated version of the breast cancer cell line BG‐1 from the NIEHS plus the genotype of a recently authenticated sample of BG‐1 from France. The complete set of results are in Supporting Information Table S2. The three match scores were calculated using CLASTR algorithms (Tanabe [TNB], Masters versus Query [MvQ], and Masters versus Reference [MvR]). The alleles at CSF1PO, D3S1358, and D5S818 that are listed in italics are less frequent alleles. The FGA locus is present as three different alleles, 23, 24, and 25, that appear in various combinations. The allelic changes at these four loci differ by one tetranucleotide repeat, suggesting that they could have arisen by errors in DNA mismatch repair. The three cell lines hTERT‐EEC‐A15‐1, ECC‐1 passage 146, and KPL‐1 (CVCL_2094) are three cell lines that are actually MCF‐7 due to lack of authentication of the starting cell line, cross‐contamination, or mislabeling. BG1 (NIEHS) is an example of a cell line taken over by MCF‐7. Its genotype differs from that of BG‐1 (France), a recently discovered sample of the original cell line. As shown in Supporting Information Table S2, the TNB scores for the 90 STR genotypes ranged between 91.3% and 100%. When the three search algorithms were used to compare the genotypes in Table 3, the scores ranged between 88.9% and 100%. Evidently, the TNB, MvR, and MvQ scoring algorithms are all sufficiently robust to identify related strains of the MCF‐7 (CVCL_0031) cell line. Table 3 and Supporting Information S2 also compare the STR profiles of several cell lines that were found by STR genotyping to have been cross‐contaminated with MCF‐7: three samples of the breast cancer cell line BG‐1 obtained from the U.S. National Institute of Environmental Health Sciences (NIEHS) and two samples of the breast cancer cell line KPL‐1 (CVCL_2094). They had been cross‐contaminated with and taken over by MCF‐7 (Korch et al., 2012; Li et al., 2014; this report). However, two profiles of BG‐1 cells from a French laboratory have since been shown to be authentic samples of the BG‐1 cell line (see CVCL_6570; Li et al., 2014). In addition, the profiles of all seven samples of the immortalized purported endometrial cell line hTERT‐EEC clones (CVCL_X007; Hombach‐Klonisch et al., 2005; Willing et al., 2011, 2013) and one sample of the endometrial Ishikawa derivative ECC‐1 (CVCL_7260) were found to have been cross‐contaminated and taken over by the breast carcinoma cell line MCF‐7 (Korch et al., 2012). Table 4 (based on data from Supporting Information Table S2) illustrates the allelic variations observed at different STR loci in the STR profiles of the different isolates of the breast cancer cell line MCF‐7 (CVCL_0031). The STR loci D5S818, D7S820, and FGA show the most variability, and yet the TNB formula readily identifies both actual MCF‐7 cell lines and lines that have been taken over by MCF‐7. Table 4. | MCF‐7 STR genotype frequencies | ||||||| |---|---|---|---|---|---|---|---| | STR locus | MCF‐7 genotype 1 | MCF‐7 genotype 2 | MCF‐7 genotype 3 | MCF‐7 genotype 4 | Genotype ratios (1:2:3:4) | Total number profiles | Most common genotype | | CSF1PO | 10 | 10,11 | 9,10 | – | 80:4:2 | 86 | 10 | | D2S1338 | 21,23 | 21, 24 | – | – | 53:1 | 54 | 21,23 | | D3S1358 | 16 | 16,17 | – | – | 75:2 | 77 | 16 | | D5S818 | 11, 12 | 12 | 11,12,13 | – | 67:16:3 | 86 | 11,12 | | D7S820 | 8,9 | 9 | 8, 9, 10 | 9,10 | 68:16:3:1 | 88 | 8,9 | | D8S1179 | 10,14 | 10,14.1 | 10,13 | – | 80:1:1 | 82 | 10,14 | | D13S317 | 11 | 10,11 | – | – | 86:1 | 87 | 11 | | D16S539 | 11,12 | – | – | – | 86 | 86 | 11,12 | | D18S51 | 14 | 14,21 | – | – | 81:1 | 82 | 14 | | D19S433 | 13,14 | – | – | – | 58 | 58 | 13,14 | | D21S11 | 30 | – | – | – | 82 | 82 | 30 | | FGA | 23,25 | 23,24,25 | 23,24 | 19,23,25 | 57:18:2:4 | 81 | 23,25 | | TH01 | 6 | – | – | – | 90 | 90 | 6 | | TPOX | 9,12 | – | – | – | 86 | 86 | 9,12 | | vWA | 14,15 | – | – | – | 90 | 90 | 14,15 | Includes STR genotypes of both samples labeled as MCF‐7 and of the imposter cell lines that have been cross‐contaminated and taken over by MCF‐7 (data from Supporting Information Table S2). Clearly, related cell lines can be identified using the TNB algorithm, as is illustrated by the data for HeLa, DLD‐1, and MCF‐7 cell lines and their derivatives (Tables 1, 2, and 3 and Supporting Information Tables S1 and S2). When this scoring method is used as described by the U.S. National Library of Medicine (NLM) and ATCC standards (Almeida & Korch, 2023; Korch et al., 2022), it meets the desired Goals 1, 2, 3, 6, 7, and 8 as presented herein and discussed by others (Capes‐Davis et al., 2013). Masters versus Reference (MvR) Algorithm However, the TNB algorithm often does not meet Goal 4 for mixed samples and Goal 5 for samples showing extensive genetic drift and low or extensive loss of heterozygosity. In the case of mixed cell cultures and cell lines that are MSI unstable, the scores are often below the recommended cutoff similarity value of 80% (Capes‐Davis et al., 2013; Korch et al., 2022; Masters et al., 2001) or the more reliable cutoff score of 90% (Bady et al., 2012; Capes‐Davis et al., 2013; Liang‐Chu et al., 2015). This is because there are often more than two peaks at several STR loci (Korch et al., 2022) in the query profile, thus increasing the denominator in the TNB formula without increasing the number of shared loci, which results in reduced apparent similarity between the two samples. The two variants of the Masters scoring algorithm produce different results when comparing two STR profiles because the denominator in the equation will be different if the two cell lines have different number of alleles at shared loci. Thus, unlike the TNB‐Sørensen algorithm, the two Masters algorithms do not satisfy Goal 3. On examining why none of these three formulae meet all eight goals, it was realized that the formulae have an underappreciated property. In the TNB formula, the denominator is the average of the number of alleles in the two cell lines, whereas in the two variants of the Masters equations, MvR and MvQ, the denominator is a single value, either the number of alleles in the reference or those in the query profile at shared STR loci. Clearly, therefore, the sample with the smallest number of alleles at shared loci limits the potential number of shared alleles. Thus, when the numbers of alleles in the two cell lines are quite different (e.g., when analyzing cultures that are mixed, are MSI unstable, or show great differences in heterozygosity), the cell line with the fewest alleles limits the potential number of shared alleles and thus Goals 3, 4, and 5 cannot be met. This realization led to the hypothesis that these three formulae (see above) would be more useful when comparing three different types of STR genotyping data: Situation 1: When comparing STR profiles with similar number of alleles at shared loci and where none of the loci have more than two alleles, one should use the TNB algorithm; Situation 2: When the query profile has a significantly higher number of alleles than the reference sample at shared loci, possibly because the query is a mixed culture of different cells or appears to be MSI unstable, one should use the MvR algorithm; and Situation 3: When the query profile is monoallelic at most or all STR loci, one should use the MvQ algorithm. Testing the Situation 1 hypothesis using the TNB algorithm and multiple MCF‐7 genotypes Tests of these situational hypotheses were performed as follows. Situation 1 is the most commonly encountered and has been commonly used in the published literature, with several studies having tested this hypothesis (Bady et al., 2012; Capes‐Davis et al., 2013; Yu et al., 2015). As has been observed, cell lines can often undergo genetic drift through in vitro cell culture, resulting in slightly different STR profiles between laboratories and between samples of the same cultures taken at different passages. From Table 3 and Supporting Information Table S2, with the 90 profiles of the breast cancer cell line MCF‐7 (CVCL_0031), and Table 4, with a summary of the frequencies of various STR genotypes of this cell line, it can be seen that even though these genotypes vary slightly, the TNB algorithm yielded scores between 91.3% and 100%, showing that it can still be used to confirm that all of these samples originated from the same donor. Cellosaurus data show that MCF‐7 has frequently been used, with 708 variants listed therein. This could explain, in part, why it has cross‐contaminated other cell lines: that is, many labs use it, and because it grows relatively quickly (doubling time ≈ 24 hr), it can take over slower‐growing cultures should it infect them. Testing the Situation 2 hypothesis using artificially combined STR genotypes The hypothesis of Situation 2 was tested by making 28 pairs of combinations of STR genotypes from Cellosaurus to create artificially mixed profiles. These “mixed” profiles (Table 5) were then analyzed by using the CLASTR program of Cellosaurus and the three analysis options: TNB, MvR, and MvQ. The results are in Supporting Information Table S4, and three selected examples are presented in Table 5. Unlike the other two algorithms, the MvR algorithm identified the individual cell line profiles used to create the artificial mixtures with match scores of 100% for all 28 pairs shown in Supporting Information Tables S3 and S4. Table 5. | Name | No. shared loci | Score (%) | Amel | CSF1PO | D2S1338 | D3S1358 | D5S818 | D7S820 | D8S1179 | D13S317 | D16S539 | D18S51 | D19S433 | D21S11 | FGA | TH01 | TPOX | vWA | | |---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---| | Mix 3 | MDA‐MB‐453, KTC‐2 Mix | X | 10,12,13 | 15 | 11,12 | 10,11 | 10,12 | 8,12 | 9,10,11 | 15,20,16 | 29,31,30 | 18,23 | 6 | 10,8 | 17,18 | |||| | TNB | MDA‐MB‐453 | 13 | 80.85 | X | 10,12 | 15 | 11 | 10 | 10,12 | 12 | 9 | 15,20 | 29,31 | 18,23 | 6 | 10 | 17,18 | || | TNB | PL517 | 8 | 77.42 | X,Y | 12 | 11,12 | 10,11 | 8,12 | 10,11 | 7,9 | 8 | 17,18 | ||||||| | TNB | MDA‐kb2 | 8 | 74.07 | X | 10,12 | 11 | 10 | 12 | 9 | 6 | 10 | 17,18 | ||||||| | TNB | KTC‐2 | 9 | 73.33 | X | 15 | 12 | 11 | 12 | 8,12 | 16 | 30 | 23 | 17,18 | |||||| | MvQ | CEM/MX1 | 8 | 70.59 | X | 10,11,13,14 | 11,12 | 9,11,12,13 | 8,11,12 | 10,11,13 | 6,7,8 | 8,11 | 16,18,19 | ||||||| | MvQ | JMAR | 8 | 70.59 | X | 10,12,13 | 10,12,13 | 9,10,12 | 8,11,12 | 8,9,12,13 | 6,8 | 8,10,11 | 15,16,18 | ||||||| | MvQ | PL517 | 8 | 70.59 | X,Y | 12 | 11,12 | 10,11 | 8,12 | 10,11 | 7,9 | 8 | 17,18 | ||||||| | MvR | MDA‐MB‐453 | 13 | 100.00 | X | 10,12 | 15 | 11 | 10 | 10,12 | 12 | 9 | 15,20 | 29,31 | 18,23 | 6 | 10 | 17,18 | || | MvR | KTC‐2 | 9 | 100.00 | X | 15 | 12 | 11 | 12 | 8,12 | 16 | 30 | 23 | 17,18 | |||||| | MvR | MDA‐kb2 | 8 | 100.00 | X | 10,12 | 11 | 10 | 12 | 9 | 6 | 10 | 17,18 | ||||||| | Mix 16 | K562, SK‐RC‐28 mix | X | 9,10,12 | 16,15 | 12,10,11 | 9,10,11 | 12,13,14 | 8,11,14 | 11,12 | 15,16 | 30,31,28,29 | 21,24,22,23 | 9.3,6,10 | 9,8,10 | 16,17,18 | |||| | TNB | SK‐RC‐28 | 13 | 77.42 | X | 10,12 | 17,24 | 15 | 10,11 | 10,11 | 12,13,14 | 11,14 | 12 | 15 | 14,15.2 | 28,29 | 22,23 | 6,10 | 8,10 | 17,18 | | TNB | K562/MTX‐2 | 8 | 75.68 | X | 9,10 | 11,12 | 9,11 | 8 | 11,12 | 6,9.3 | 8,9 | 16 | ||||||| | TNB | K‐562 | 13 | 73.33 | X | 9,10 | 17 | 16 | 11,12 | 9,11 | 12 | 8 | 11,12 | 15,16 | 14,14.2 | 29,30,31 | 21,24 | 9.3 | 8,9 | 16 | | MvQ | Fsp 62891 | 8 | 65.22 | X,Y | 7,10,12,13 | 12 | 8,9,11,12 | 8,11,14 | 11,12,13 | 6,7,8,9.3 | 8,9,11 | 16,19 | ||||||| | MvQ | IST‐1 | 8 | 65.22 | X | 10,11,12 | 11,12 | 9,10,13,14 | 8,11,12 | 10,12,13 | 7,8,9,9.3 | 6,8,9,11 | 16,17,18 | ||||||| | MvQ | SVG p12 | 8 | 65.22 | X,Y | 10,11,12 | 11,12,14 | 10,12,13 | 8,13,14 | 10,11,12 | 6,7,9,9.3 | 8,10,11 | 15,16,18,19 | ||||||| | MvR | K‐562 | 13 | 100.00 | X | 9,10 | 17 | 16 | 11,12 | 9,11 | 12 | 8 | 11,12 | 15 | 14,14.2 | 29,30 | 21,24 | 9.3 | 8,9 | 16 | | MvR | K‐562 | 13 | 90.91 | X | 9,10 | 17 | 16 | 11,12,13 | 9,11 | 12 | 8 | 11,12,13 | 15 | 14,14.2 | 29,30 | 21,24 | 9.3 | 8,9 | 16 | | MvR | SK‐RC‐28 | 13 | 100.00 | X | 10,12 | 17,24 | 15 | 10,11 | 10,11 | 12,13,14 | 11,14 | 12 | 15 | 14,15.2 | 28,29 | 22,23 | 6,10 | 8,10 | 17,18 | | Mix 17 | MDA‐MB‐453, OVTOKO Mix | X | 10,12 | 15 | 11,12 | 9,10 | 10,12,14 | 8,12 | 9,11,12 | 15,20,19 | 29,31 | 18,21,23 | 6,9.3 | 10,8,12 | 17,18,14 | |||| | TNB | OVTOKO | 13 | 85.19 | X | 12 | 15 | 11,12 | 9,10 | 10,14 | 8,12 | 11,12 | 19 | 29,31 | 21,23 | 6,9.3 | 8,12 | 14,17 | || | TNB | MDA‐MB‐453 | 13 | 76.00 | X | 10,12 | 15 | 11 | 10 | 10,12 | 12 | 9 | 15,20 | 29,31 | 18,23 | 6 | 10 | 17,18 | || | TNB | NCI‐H1284 | 8 | 74.29 | X,Y | 10,12 | 11,12 | 9,10,11 | 8,12 | 12,13 | 9.3 | 8,10 | 17,19 | ||||||| | MvQ | OVTOKO | 13 | 74.19 | X | 12 | 15 | 11,12 | 9,10 | 10,14 | 8,12 | 11,12 | 19 | 29,31 | 21,23 | 6,9.3 | 8,12 | 14,17 | || | MvQ | CLNH11.4 | 8 | 68.42 | X,Y | 10,11,12 | 10,11,12,13 | 9,11 | 11 | 9,11,12 | 6,7,8,9.3 | 8,11 | 14,17,19,20 | ||||||| | MvQ | IST‐1 | 8 | 68.42 | X | 10,11,12 | 11,12 | 9,10,13,14 | 8,11,12 | 10,12,13 | 7,8,9,9.3 | 6,8,9,11 | 16,17,18 | ||||||| | MvR | MDA‐MB‐453 | 13 | 100.00 | X | 10,12 | 15 | 11 | 10 | 10,12 | 12 | 9 | 15,20 | 29,31 | 18,23 | 6 | 10 | 17,18 | || | MvR | OVTOKO | 13 | 100.00 | X | 12 | 15 | 11,12 | 9,10 | 10,14 | 8,12 | 11,12 | 19 | 29,31 | 21,23 | 6,9.3 | 8,12 | 14,17 | || | MvR | MDA‐kb2 | 8 | 100.00 | X | 10,12 | 11 | 10 | 12 | 9 | 6 | 10 | 17,18 | Comparison of match scores for three different pairs of cell line genotypes combined artificially to determine which algorithm can identify the cell lines in mixtures. Algorithms: TNB, Tanabe; MvQ, Masters versus Query; MvR, Masters versus Reference. Only the top three or four resulting cell line genotypes that are unique were selected for illustration purposes from Supporting Information Table S3, first sheet, which summarizes the complete results shown in the second through fourth sheets of the same table. The alleles in red font are those that are discordant between the query and reference genotypes or missing from the genotype (e.g., at D2S1338, Penta D, Penta E loci). Note that only the MvR algorithm allowed identification of the cell lines in the mixtures. Situation 2 was further tested by artificially creating two triple mixtures of STR genotypes using the following profiles: Mixture A: CaLu‐6 (CVCL_0238) + T241 (CVCL_M975) + NCI‐H838 (CVCL_1594); and Mixture B: NCI‐H838 (CVCL_1594) + OVTOKO (CVCL_3117) + SK‐RC‐28 (CVCL_6184). In Table 6 (Mixture A), the TNB and MvQ algorithms yielded scores of ≤80%, whereas the MvR scores were 100% for the three components plus for two cell lines that have the same profile as CaLu‐6, that is, Caov‐2 (CVCL_6861) and OVCAR‐2 (CVCL_3941). In 2025, the OVCAR‐2 reference profile in Cellosaurus was shown to be incorrect based on testing of an original sample of this cell line (Kerry Campbell, Fox Chase Cancer Center, personal communication). In Table 6 (Mixture B), the TNB and MvQ scores were <74%, whereas the three component profiles produced MvR scores of 100%. Table 6. | Names | Cellosaurus ID | No. shared loci | Score (%) | Amel | CSF1PO | D2S1338 | D3S1358 | D5S818 | D7S820 | D8S1179 | D13S317 | D16S539 | D18S51 | D19S433 | D21S11 | FGA | Penta D | Penta E | TH01 | TPOX | vWA | | |---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---| | Calu‐6 a | CVCL_0238 | na | na | X | 12 | 17,26 | 16 | 11 | 10 | 10,14 | 11 | 13 | 12,16 | 13 | 31 | 22 | 13 | 5,14 | 9 | 8 | 17 | | | T241 a | CVCL_M975 | na | na | X | 12 | 15 | 13 | 11 | 12 | 12 | 9 | 19 | 30 | 21 | 9.3 | 8 | 15 | ||||| | NCI‐H838 a | CVCL_1594 | na | na | X | 10 | 18 | 15 | 11 | 9 | 13,16 | 8 | 9,12 | 12 | 15.2 | 32.2 | 22.2 | 11 | 5,12 | 7 | 8,11 | 17 | | | Mixture A: Calu‐6 + T241 + NCI‐H838 a | Query profile | X | 12,10 | 17,26,18 | 16,15 | 11,13 | 10,11,9 | 10,14,12, 13,16, | 11,12,8 | 13,9,12 | 12,16,19 | 13,15.2 | 31,30,32.2 | 22,21,22.2 | 13,11 | 514,12 | 9,9.3,7 | 8,11 | 15,17 | ||| | TNB | GM00232 | CVCL_Y965 | 8 | 74.29 | X,Y | 10,12 | 13 | 10,11 | 11,12 | 11,13 | 8,9.3 | 8,11 | 15,17 | ||||||||| | TNB | Hs 728.Sk/Mu | CVCL_0870 | 8 | 74.29 | X,Y | 10,11 | 7,13 | 9,10 | 11,12 | 12,13 | 7,9.3 | 8 | 15,17 | ||||||||| | TNB | Hs 728.Sp | CVCL_1044 | 8 | 74.29 | X,Y | 10,11 | 7,13 | 9,10 | 11,12 | 12,13 | 7,9.3 | 8 | 15,17 | ||||||||| | TNB | CG0446 | CVCL_LG59 | 8 | 72.73 | X,Y | 10,12 | 11,13 | 10,11 | 9,12 | 9,12 | 9 | 8 | 17 | ||||||||| | TNB | CX16‐RV5 Ecto | CVCL_B0AE | 8 | 72.22 | X | 10,12 | 11,13 | 9,10 | 11,12 | 10,13 | 7,8 | 8,11 | 17,18 | ||||||||| | MvQ | IST‐1 b | CVCL_F642 | 8 | 80.00 | X | 10,11,12 | 11,12 | 9,10,13,14 | 8,11,12 | 10,12,13 | 7,8,9,9.3 | 6,8,9,11 | 16,17,18 | ||||||||| | MvQ | DM14 c | CVCL_1Q84 | 8 | 75.00 | X | 10,11,12,14 | 9,11,13 | 8,9,12 | 8,11,13 | 9,12,13 | 6,7 | 8,11 | 15,17,18,20 | ||||||||| | MvQ | DM12 c | CVCL_1Q83 | 8 | 70.00 | X | 10,11,12,14 | 9,11,13 | 8,9,12 | 8,11,13 | 9,12,13 | 6,7 | 8,11 | 15,18,20 | ||||||||| | MvQ | JMAR | CVCL_1Q86 | 8 | 70.00 | X | 10,12,13 | 10,12,13 | 9,10,12 | 8,11,12 | 8,9,12,13 | 6,8 | 8,10,11 | 15,16,18 | ||||||||| | MvR | Calu‐6 | CVCL_0236 | 17 | 100.00 | X | 12 | 17,26 | 16 | 11 | 10 | 10,14 | 11 | 13 | 12,16 | 13 | 31 | 22 | 13 | 5,14 | 9 | 8 | 17 | | MvR | Caov‐2 | CVCL_6861 | 15 | 100.00 | X | 12 | 17,26 | 16 | 11 | 10 | 10,14 | 11 | 13 | 12,16 | 13 | 31 | 22 | 9 | 8 | 17 | || | MvR | OVCAR‐2 | CVCL_3941 | 15 | 100.00 | X | 12 | 17,26 | 16 | 11 | 10 | 10,14 | 11 | 13 | 12,16 | 13 | 31 | 22 | 9 | 8 | 17 | || | MvR | T241 | CVCL_M975 | 13 | 100.00 | X | 12 | 15 | 13 | 11 | 12 | 12 | 9 | 19 | 30 | 21 | 9.3 | 8 | 15 | |||| | MvR | NCI‐H838 | CVCL_1594 | 17 | 100.00 | X | 10 | 18 | 15 | 11 | 9 | 13,16 | 8 | 9,12 | 12 | 15.2 | 32.2 | 22.2 | 11 | 5,12 | 7 | 8,11 | 17 | | MvR | HCC1195 | CVCL_5127 | 10 | 100.00 | X,Y | 12 | 15,16 | 13 | 10,11 | 12 | 9 | 21 | 7 | 8,11 | 15,17 | ||||||| | NCI‐H838 a | CVCL_1594 | na | na | X | 10 | 18 | 15 | 11 | 9 | 13,16 | 8 | 9,12 | 12 | 15.2 | 32.2 | 22.2 | 11 | 5,12 | 7 | 8,11 | 17 | | | OVTOKO a | CVCL_3117 | na | na | X | 12 | 15 | 11,12 | 9,10 | 10,14 | 8,12 | 11,12 | 19 | 29,31 | 21,23 | 9,10 | 14 | 6,9.3 | 8,12 | 14,17 | ||| | SK‐RC‐28 a | CVCL_6184 | na | na | X | 10,12 | 17,24 | 15 | 10,11 | 10,11 | 12,13,14 | 11,14 | 12 | 15 | 14,15.2 | 28,29 | 22,23 | 6,10 | 8,10 | 17,18 | ||| | Mixture B: NCI‐H838 + OVTOKO + SK‐RC‐28 a | Query profile | X | 10,12 | 18,17,24 | 15 | 11,12,10 | 9,10,11 | 13,16,10,14,12 | 8,12,14,11 | 9,12,11 | 12,19,15 | 15.2,14 | 32.2,29,31,28 | 22.2,21,23,22 | 11,9,10 | 5,12,14 | 7,6,9.3,10 | 8,11,12,10 | 17,14,18 | ||| | TNB | SK‐RC‐28 | CVCL_6184 | 15 | 73.68 | X | 10,12 | 17,24 | 15 | 10,11 | 10,11 | 12,13,14 | 11,14 | 12 | 15 | 14,15.2 | 28,29 | 22,23 | 6,10 | 8,10 | 17,18 | || | TNB | OVTOKO | CVCL_3117 | 15 | 69.33 | X | 12 | 15 | 11,12 | 9,10 | 10,14 | 8,12 | 11,12 | 19 | 29,31 | 21,23 | 9,10 | 14 | 6,9.3 | 8,12 | 14,17 | || | TNB | AP78P | CVCL_JH2O | 8 | 68.29 | X,Y | 10,12 | 10,11 | 10 | 8,11 | 11,12 | 6,9 | 8,10 | 14,18 | ||||||||| | TNB | D422T | CVCL_2883 | 8 | 68.29 | X,Y | 10,12 | 11,12 | 9,12 | 12 | 11,12 | 7,9.3 | 8,11 | 14,17 | ||||||||| | MvQ | IST‐1 a | CVCL_F642 | 8 | 61.54 | X | 10,11,12 | 11,12 | 9,10,13,14 | 8,11,12 | 10,12,13 | 7,8,9,9.3 | 6,8,9,11 | 16,17,18 | ||||||||| | MvQ | JMAR | CVCL_1Q86 | 8 | 61.54 | X | 10,12,13 | 10,12,13 | 9,10,12 | 8,11,12 | 8,9,12,13 | 6,8 | 8,10,11 | 15,16,18 | ||||||||| | MvR | NCI‐H838 | CVCL_1594 | 17 | 100.00 | X | 10 | 18 | 15 | 11 | 9 | 13,16 | 8 | 9,12 | 12 | 15.2 | 32.2 | 22.2 | 11 | 5,12 | 7 | 8,11 | 17 | | MvR | OVTOKO | CVCL_3117 | 15 | 100.00 | X | 12 | 15 | 11,12 | 9,10 | 10,14 | 8,12 | 11,12 | 19 | 29,31 | 21,23 | 9,10 | 14 | 6,9.3 | 8,12 | 14,17 | || | MvR | SK‐RC‐28 | CVCL_6184 | 15 | 100.00 | X | 10,12 | 17,24 | 15 | 10,11 | 10,11 | 12,13,14 | 11,14 | 12 | 15 | 14,15.2 | 28,29 | 22,23 | 6,10 | 8,10 | 17,18 | || | MvR | HECV | CVCL_0295 | 9 | 100.00 | X | 12 | 10 | 10,11 | 12 | 9 | 29 | 6 | 8,11 | 17 | |||||||| | MvR | LM‐MEL‐42 | CVCL_UC48 | 9 | 100.00 | X,Y | 10,12 | 10,12 | 10,11 | 11 | 12 | 29 | 6,9.3 | 8 | 17,18 | |||||||| | MvR | SLR20 | CVCL_V606 | 9 | 100.00 | X | 10,12 | 10,12 | 10,11 | 12 | 9 | 29 | 6 | 8,11 | 17 | |||||||| | MvR | EDi022‐A | CVCL_VT08 | 8 | 100.00 | X,Y | 10 | 11,12 | 9,10 | 12,14 | 11,12 | 9.3 | 8 | 18 | ||||||||| | MvR | GSC#112 | CVCL_A9TN | 8 | 100.00 | X | 10 | 12 | 11 | 12 | 11 | 7 | 11 | 17 | ||||||||| | MvR | HBMEC‐2 | CVCL_AT91 | 8 | 100.00 | X | 12 | 10 | 10,11 | 12 | 9 | 6 | 8,11 | 17 | ||||||||| | MvR | Hs 80.Fs | CVCL_0918 | 8 | 100.00 | X,Y | 10,12 | 11,12 | 10 | 11,12 | 9,12 | 7 | 11 | 14,18 | ||||||||| | MvR | PCI‐22B | CVCL_1T57 | 8 | 100.00 | X | 10,12 | 10,12 | 10,11 | 12 | 9 | 6 | 8,11 | 17 | ||||||||| | MvR | SKR1 | CVCL_G366 | 8 | 100.00 | X | 10,12 | 10,12 | 10,11 | 12 | 9 | 6 | 8,11 | 17 | ||||||||| | MvR | SNK‐57 | CVCL_3923 | 8 | 100.00 | X,Y | 12 | 12 | 10,11 | 11,12 | 9,12 | 7 | 8 | 17 | ||||||||| | MvR | UC702 | CVCL_3226 | 8 | 100.00 | X | 12 | 11 | 11 | 11 | 11,12 | 6,7 | 8 | 14,17 | ||||||||| | MvR | VMM17 | CVCL_S581 | 8 | 100.00 | X | 10,12 | 12 | 9,10 | 11,12 | 12 | 9.3 | 8,11 | 14,17 |

profiles. IST‐1: human contaminated placenta cell line shown to be actually a SK‐OV‐3 (CVC_L0532) derivative. DM12, DM24: human oral cavity squamous cell carcinoma (OCSCC). Two mixtures of three cell line STR genotypes—Mixture A (Calu‐6 + T241 + NCI‐H838) and Mixture B (NCI‐H838 + OVTOKO + SK‐RC‐28)—were combined in silico to make artificial genotypes that were used to search for matches using the CLASTR search tool and the three different search algorithms—Tanabe (TNB), Masters versus Query (MvQ), and Masters versus Reference (MvR). The alleles in red font are those that are discordant between the query and reference genotypes or missing from the genotype (e.g., at the D2S1338, Penta D, and Penta E loci). Only the MvR algorithm identified the cell line genotypes with 100% scores. Mixture A did identify one unrelated cell line (HCC1195, CVCL_5127) with a 100% match scores, and Mixture B identified 12 unrelated cell lines with 100% match scores. Note that in both cases, the genotypes of the unrelated cell lines were determined by analysis at 10 or 8 STR loci, whereas the true matches were genotyped at 13, 15, or 17 loci, emphasizing that it is best to STR genotype at as many loci as possible. One additional cell was identified as a 100% match to the mixture with the MvR algorithm (HCC1195; CVCL_5127), which is from a lung adenosquamous carcinoma. It is not related to any of the individual components. Using Mixture B, several other cell lines also gave MvR scores of 100%. The identification of these unrelated matches to the triple profile mixtures is due to the fact that there are several possible allele combinations in these triple profile mixtures. These false positive cell lines would need to be checked for additional genetic markers, for example by using an STR kit that tests additional loci or by confirmation of SNPs specific for one of more components in the cell lines. Testing the Situation 2 hypothesis further, the MvR algorithm was applied to cell line samples that showed more than two alleles at several STR loci. Cell line profiles can show more than two alleles at multiple STR loci for three possible reasons: (1) because the cell line culture is a mixture of two or more distinct cell lines, (2) because the cell line is genetically unstable due to defects in the DNA MMR system, or (3) because the cell line has arisen through the fusion of two different cell lines. In the first instance, such profiles will clearly be a mixture of two cell lines, which is illustrated by the above‐described results for the artificially mixed cell line profiles (Tables 5 and 6 and Supporting Information Table S3). The purported thyroid cell line BHP 18‐21 (CVCL_6282) was observed to have several loci with three or four alleles (peaks) in its profile (Nims et al., 2018; Schweppe et al., 2008). The match scores obtained using the three algorithms are shown in Table 7 (data from Supporting Information Table S3, sample 29, plus additional data not shown). The MvR algorithm scores were 100%, compared to scores of <81% for TNB and MvQ. This indicates that this cell line is a mixture of the male‐melanoma‐derived cell line M14 (CVCL_1395) and the thyroid cancer cell line TPC‐1 (CVCL_6298). Furthermore, on searching for matches to those two cell lines, several other cell lines purportedly derived from thyroid tissue were identified that had been shown earlier to be cross‐contaminated and taken over by either M14 or TPC‐1 (Schweppe et al., 2008). As described for MCF‐7 above, a minor amount of genetic drift can lead some cultures to lose an STR allele at one or a few loci between passages, as shown in Supporting Information Tables S2 and S3. Table 7. | Name | ID | No. shared STR loci | Score (%) | Amel | CSF1PO | D2S1338 | D3S1358 | D5S818 | D7S820 | D8S1179 | D13S317 | D16S539 | D18S51 | D19S433 | D21S11 | FGA | Penta D | Penta E | TH01 | TPOX | vWA | | |---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---| | Query STR | BHP 18‐21 | DP0047‐07 | 9 | na | X | 14,16,17 | 8,10,11,12 | 8,10,11 | 11,13,17 | 11,12 | 13,16,17 | 30,31.2 | 20,21 | 14,16,18 | |||||||| | TNB | BHP 18‐21 | CVCL_6282 | 9 | 100.00% | X | 14,16,17 | 8,10,11,13 | 8,10,11 | 11,13,17 | 11,12 | 13,16,17 | 30,31.2 | 20,21 | 14,16,18 | |||||||| | TNB | BHP 2‐7 | CVCL_6283 | 8 | 81.08% | X | 11,12 | 16,17 | 8,10 | 11 | 11,12 | 9 | 13,16 | 30,31.2 | 20,21 | 9 | 11 | 14,18 | ||||| | TNB | BHP 10‐3 | CVCL_6278 | 9 | 80.95% | X | 16,17 | 8,10 | 11 | 11,17 | 11,12 | 13,16 | 30,31.2 | 20,21 | 14,18 | |||||||| | TNB | BHP 7‐13 | CVCL_6285 | 9 | 80.95% | X | 16,17 | 8,10 | 11 | 11,17 | 11,12 | 13,16 | 30,31.2 | 20,21 | 14,18 | |||||||| | TNB | TPC‐1 | CVCL_6298 | 9 | 80.95% | X | 11,12 | 16,17 | 8,10 | 11 | 11,17 | 11,12 | 9 | 13,16 | 30,31.2 | 20,21 | 9,13 | 18 | 9 | 11 | 14,18 | || | MvQ | BHP 18‐21 | CVCL_6282 | 9 | 100.00% | X | 14,16,17 | 8,10,11,13 | 8,10,11 | 11,13,17 | 11,12 | 13,16,17 | 30,31.2 | 20,21 | 14,16,18 | |||||||| | MvQ | BHP 2‐7 | CVCL_6283 | 8 | 68.18% | X | 8,10 | 16,17 | 8,10 | 9,11 | 11,12 | 9 | 13,16 | 30,31.2 | 20,21 | 11 | 11 | 14,18 | ||||| | MvQ | BHP 10‐3 | CVCL_6278 | 9 | 68.00% | X | 16,17 | 8,10 | 11 | 11,17 | 11,12 | 13,16 | 30,31.2 | 20,21 | 14,18 | |||||||| | MvQ | BHP 7‐13 | CVCL_6285 | 9 | 68.00% | X | 16,17 | 8,10 | 11 | 11,17 | 11,12 | 13,16 | 30,31.2 | 20,21 | 14,18 | |||||||| | MvQ | TPC‐1 | CVCL_6298 | 9 | 68.00% | X | 11,12 | 16,17 | 8,10 | 11 | 11,17 | 11,12 | 9 | 13,16 | 30,31.2 | 20,21 | 9,13 | 18 | 9 | 11 | 14,18 | || | MvR | M14 | Reference | 15 | 92.86% | X | 11 | 19,24 | 14,16 | 11,12 | 8,10 | 13 | 12 | 9,13 | 13,17 | 14,15 | 30 | 21 | 6,7 | 8,11 | 16,18 | || | MvR | M14 | CVCL_1395 | 9 | 92.86% | X | 11 | 19,24 | 14,16 | 11,12 | 8,10 | 13 | 12 | 9,13 | 13,17 | 14,15 | 30 | 21 | 6,7 | 8,11 | 16,18 | || | MvR | BHP 18‐21 | CVCL_6282 | 9 | 100.00% | X | 14,16,17 | 8,10,11,13 | 8,10,11 | 11,13,17 | 11,12 | 13,16,17 | 30,31.2 | 20,21 | 14,16,18 | |||||||| | MvR | BHP 14‐9 | CVCL_6279 | 9 | 92.86% | X | 14,16 | 11,12 | 8,10 | 13 | 12 | 13,17 | 30 | 21 | 16,18 | |||||||| | MvR | BHP 15‐3 | CVCL_6280 | 9 | 92.86% | X | 14,16 | 11,12 | 8,10 | 13 | 12 | 13,17 | 30 | 21 | 16,18 | |||||||| | MvR | BHP 17‐10 | CVCL_6281 | 9 | 92.86% | X | 14,16 | 11,12 | 8,10 | 13 | 12 | 13,17 | 30 | 21 | 16,18 | |||||||| | MvR | BHP 5‐16 | CVCL_6284 | 9 | 92.86% | X | 11 | 14,16 | 11,12 | 8,10 | 13 | 12 | 9,13 | 13,17 | 30 | 21 | 6,7 | 8,11 | 16,18 | |||| | MvR | MDA‐MB‐435 | CVCL_0417 | 9 | 92.31% | X | 11 | 19,24 | 14 | 11,12 | 8,10 | 13 | 12 | 13 | 13,17 | 14 | 30 | 21 | 9,11 | 10,12 | 6,7 | 8,11 | 16,18 | | MvR | MDA‐MB‐435S | CVCL_0622 | 9 | 92.31% | X | 11 | 19,24 | 14 | 11,12 | 8,10 | 13 | 12 | 13 | 13,17 | 14 | 30 | 21 | 9,11 | 10,12 | 6,7 | 8,11 | 16,18 | | MvR | MDA‐N | CVCL_1910 | 9 | 92.31% | X | 11 | 19,24 | 14 | 11,12 | 8,10 | 13 | 12 | 13 | 13,17 | 14 | 30 | 21 | 6,7 | 8,11 | 16,18 | || | MvR | NPA | CVCL_0467 | 9 | 92.31% | X | 11 | 14,16 | 11,12 | 8 | 13 | 12 | 9,13 | 13,17 | 30 | 21 | 6,7 | 8,11 | 16,18 | |||| | MvR | MuM‐2C | CVCL_3448 | 9 | 84.62% | X | 11 | 19 | 14,16 | 11,12 | 8,10 | 13 | 12 | 9 | 13,18 | 14,15 | 30 | 21 | 6,7 | 8,11 | 18 | || | MvR | OCM‐1 | CVCL_6934 | 9 | 84.62% | X | 11 | 14,16 | 11,12 | 8,10 | 13 | 12 | 9 | 13,18 | 30 | 21 | 6,7 | 8,11 | 18 | |||| | MvR | OCM‐1A | CVCL_6935 | 9 | 84.62% | X | 11 | 19 | 14,16 | 11,12 | 8,10 | 13 | 12 | 9 | 13,18 | 14,15 | 30 | 21 | 6,7 | 8,11 | 18 | || | MvR | TPC‐1 | Reference | 13 | 100.00% | X | 11,12 | 16,17 | 8,10 | 11 | 11,17 | 11,12 | 9 | 13,16 | 30,31.2 | 20,21 | 9 | 11 | 14,18 | |||| | MvR | TPC‐1 | CVCL_6298 | 9 | 100.00% | X | 11,12 | 16,17 | 8,10 | 11 | 11,17 | 11,12 | 9 | 13,16 | 30,31.2 | 20,21 | 9,13 | 18 | 9 | 11 | 14,18 | || | MvR | BHP 18‐21 | CVCL_6282 | 9 | 100.00% | X | 14,16,17 | 8,10,11,13 | 8,10,11 | 11,13,17 | 11,12 | 13,16,17 | 30,31.2 | 20,21 | 14,16,18 | |||||||| | MvR | BHP 10‐3 | CVCL_6278 | 9 | 100.00% | X | 16,17 | 8,10 | 11 | 11,17 | 11,12 | 13,16 | 30,31.2 | 20,21 | 14,18 | |||||||| | MvR | BHP 7‐13 | CVCL_6285 | 9 | 100.00% | X | 16,17 | 8,10 | 11 | 11,17 | 11,12 | 13,16 | 30,31.2 | 20,21 | 14,18 | |||||||| | MvR | BHP 2‐7 | CVCL_6283 | 8 | 100.00% | X | 11,12 | 16,17 | 8,10 | 11 | 11,12 | 9 | 13,16 | 30,31.2 | 20,21 | 9 | 11 | 14,18 | ||||| | MvR | BHP 7‐13 | CVCL_6285 | 9 | 100.00% | X | 16,17 | 8,10 | 11 | 11,17 | 11,12 | 13,16 | 30,31.2 | 20,21 | 14,18 | |||||||| | MvR | TPC‐1 | CVCL_6298 | 9 | 100.00% | X | 11,12 | 16,17 | 8,10 | 11 | 11,17 | 11,12 | 9 | 13,16 | 30,31.2 | 20,21 | 9,13 | 18 | 9 | 11 | 14,18 | || | MvR | BHP 2‐7 | CVCL_6283 | 8 | 100.00% | X | 11,12 | 16,17 | 8,10 | 11 | 11,12 | 9 | 13,16 | 30,31.2 | 20,21 | 9 | 11 | 14,18 | The genotype of this cell line was used to query the CLASTR database of STR genotypes for potential matches with the three algorithms: Tanabe (TNB), Masters versus Query (MvQ), and Masters versus Reference (MvR). Multiple matches to two cell lines—M14 and TPC‐1—were found. The grayed‐out allele numbers are alleles in loci of the reference genotype which are not reported for the Query profile for BHP 18‐21 (e.g., CSF1PO, D2S1338, D16S539, etc.). Cellosaurus identifiers in red font are problematic cell lines. Hovering the mouse over these red identifiers will show information describing the problem with the cell line. The STR genotype of BHP 18‐21 (CVCL_6282) illustrates how the MvR algorithm can identify the cell lines composing mixed cultures. The STR genotype in sample 30 in Table 8 and Supporting Information Table S4 is for the cell line labeled “Contaminated MDA‐MB‐453,” which is described in Figure F.4 of Appendix F of the ANSI‐ATCC ASN‐0002 Standard (Korch et al., 2022). When the three algorithms were tested, the TNB and MvQ scores were <79%, whereas the MvR scores were 100%, for authentic samples of MDA‐MB‐453 (CVCL_0418, CU‐AMC DP0165‐03, and DP0290‐07), MDA‐kb2 (CVCL_6421; a known derivative of MDA‐MB‐453), two samples of K562 (CU‐AMC DP0222‐03 and CVCL_0004), and four other potential matches to K562. Nine other cell lines had MvR scores between 91.7% and 85.7%. The identity of these nine lines might require further study to verify whether they are related to either MDA‐MB‐453 or K562. The MvR algorithm is the best one to use to identify the cell lines in mixed cultures. Table 8. | Name | No. shared loci | Score (%) | Amel | CSF1PO | D2S1338 | D3S1358 | D5S818 | D7S820 | D8S1179 | D13S317 | D16S539 | D18S51 | D19S433 | D21S11 | FGA | Penta D | Penta E | TH01 | TPOX | vWA | || |---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---| | Mix 30 | Contaminated MDA‐MB‐453 | Cellosaurus ID | X | 10,12 | 17,23,24 | 15,16 | 11,12 | 10,11 | 10,12 | 8,12 | 9,11,12 | 15,16,20 | 13,14,14.2 | 29,31 | 18,21,23,24 | 9,10,13 | 5,11,14 | 6,9.3 | 8,10 | 17,18 | || | TNB | NCI‐H1284 | CVCL_1461 | 8 | 76.47% | X,Y | 10,12 | 11,12 | 9,10,11 | 8,12 | 12,13 | 9.3 | 8,10 | 17,19 | ||||||||| | TNB | MDA‐kb2 | CVCL_6421 | 15 | 75.41% | X | 10,12 | 23,24 | 15 | 11 | 10 | 10,12 | 12 | 9 | 15,20 | 13,14 | 29,31 | 18,23 | 9,10 | 11 | 6 | 10 | 17,18 | | TNB | MDA‐MB‐453 | CVCL_0418 | 15 | 75.41% | X | 10,12 | 23,24 | 15 | 11 | 10 | 10,12 | 12 | 9 | 15,20 | 13,14 | 29,31 | 18,23 | 9,10 | 11 | 6 | 10 | 17,18 | | TNB | SV7tert | CVCL_3602 | 8 | 75.00% | X | 10,12 | 11,13 | 11 | 8 | 9,11 | 6,9.3 | 8,11 | 17,18 | ||||||||| | MvQ | IST‐1 a | CVCL_F642 | 8 | 72.22% | X | 10,11,12 | 11,12 | 9,10,13,14 | 8,11,12 | 10,12,13 | 7,8,9,9.3 | 6,8,9,11 | 16,17,18 | ||||||||| | MvQ | NCI‐H1284 | CVCL_1461 | 8 | 72.22% | X,Y | 10,12 | 11,12 | 9,10,11 | 8,12 | 12,13 | 9.3 | 8,10 | 17,19 | ||||||||| | MvQ | UABCATe001‐A | CVCL_D6JC | 9 | 66.67% | X | 10,12 | 9,12 | 8,11 | 8,13 | 11,12 | 30,31 | 6,9.3 | 8,9 | 15,18 | |||||||| | MvQ | Hs 207.T | CVCL_3519 | 8 | 66.67% | X | 10,12 | 12 | 11,12 | 11,12 | 11,12 | 6,7 | 8,10 | 17,18 | ||||||||| | MvR | MDA‐kb2 | CVCL_6421 | 15 | 100.00% | X | 10,12 | 23,24 | 15 | 11 | 10 | 10,12 | 12 | 9 | 15,20 | 13,14 | 29,31 | 18,23 | 9,10 | 11 | 6 | 10 | 17,18 | | MvR | MDA‐MB‐453 | CVCL_0418 | 15 | 100.00% | X | 10,12 | 23,24 | 15 | 11 | 10 | 10,12 | 12 | 9 | 15,20 | 13,14 | 29,31 | 18,23 | 9,10 | 11 | 6 | 10 | 17,18 | | MvR | MDA‐MB‐453 | CU‐AMC DP0165‐03 & DP0290‐07 | 15 | 100.0% | X | 10,12 | 23,24 | 15 | 11 | 10 | 10,12 | 12 | 9 | 15,20 | 13,14 | 29,31 | 18,23 | 6 | 10 | 17,18 | || | MvR | K562 | CU‐AMC DP0222‐03 | 15 | 100.0% | X | 9,10 | 17 | 16 | 11,12 | 9,11 | 12 | 8 | 11,12 | 15,16 | 14,14.2 | 29,30,31 | 21,24 | 9.3 | 8,9 | 16 | || | MvR | HROC57 | CVCL_1G03 | 9 | 91.67% | X | 10 | 11,12 | 11 | 12 | 9,12 | 30 | 6 | 8 | 14,17 | |||||||| | MvR | MMAc‐SF | CVCL_1420 | 8 | 90.00% | X | 12 | 12 | 11 | 8,12 | 9,12 | 9 | 8 | 17 | ||||||||| | MvR | MSK‐922 | CVCL_6989 | 8 | 90.00% | X | 10,12 | 11 | 11 | 12 | 9,11 | 8 | 8 | 17 | ||||||||| | MvR | NCC‐MPNST4‐C1 | CVCL_YU16 | 9 | 88.89% | X,Y | 12 | 12 | 11 | 10 | 9 | 30 | 6 | 8 | 17 | |||||||| | MvR | KTC‐2 | CVCL_6476 | 13 | 88.24% | X | 12,13 | 15 | 12 | 11 | 12 | 8,12 | 10,11 | 16 | 30 | 23 | 6 | 8 | 17,18 | |||| | MvR | K‐562 | CVCL_0004 | 15 | 88.00% | X | 9,10 | 17 | 16 | 11,12 | 9,11 | 12 | 8 | 11,12 | 15,16 | 14,14.2 | 29,30,31 | 21,24 | 9,13 | 5,14 | 9.3 | 8,9 | 16 | | MvR | PER‐703 | CVCL_IY99 | 9 | 86.67% | X | 10,12 | 12 | 10,11 | 12,13 | 9,11 | 30,31.2 | 9.3 | 8,9 | 17 | |||||||| | MvR | LM‐MEL‐42 | CVCL_UC48 | 9 | 85.71% | X,Y | 10,12 | 10,12 | 10,11 | 11 | 12 | 29 | 6,9.3 | 8 | 17,18 | |||||||| | MvR | SV7tert | CVCL_3602 | 8 | 85.71% | X | 10,12 | 11,13 | 11 | 8 | 9,11 | 6,9.3 | 8,11 | 17,18 | “30” at top left refers to the 30th mixed sample shown in Supplementary Information Table 3 (second and third sheets). Red font denotes alleles that are discordant between the query and reference genotypes or missing from the genotype (e.g., at D2S1338, Penta D, Penta E loci). The genotype has major peaks from MDA‐MB‐453 (CVCL_0418, CU‐AMC DP0165‐03, and CU‐AMC DP0290‐07) plus minor peaks belonging to K562 (CVCL_0004 and CU‐AMC DP0222‐03). Cell line STR genotypes determined by the author have the CU‐AMC DP0XXX‐XX identifiers. Algorithms: TNB, Tanabe; MvQ, Masters versus Query; MvR, Masters versus Reference. This is another example illustrating (1) the utility of the MvR algorithm for identifying component genotypes in a cell line mixture and (2) the increased reliability of matches when using genotypes based on multiple (13 or more) STR loci (i.e., fewer discordant alleles). Testing Situation 2 hypothesis—a tale of five MSI‐unstable cell lines: Ishikawa, Ishikawa 3‐H‐4, Ishikawa 3‐H‐12, EnCa101, and ECC‐1 In other instances of cell lines with multiple (two or more) alleles per locus, the cell line may manifest extensive genetic drift due to MSI; that is, at several STR loci it will have peaks that differ from the original allele in the gain or loss of one STR tetranucleotide repeat. This phenomenon has been described previously and manifests more clearly in some STR loci than in others and differently between different sublines of a cell line. The proportion of cells in the culture with the different alleles can vary over time between passages, with some alleles disappearing and others appearing or reappearing. In MCF‐7, the FGA locus has various combinations of the three alleles 23, 24, and 25; see Table 3 and Supporting Information Table S2 for the variations in the allelic genotypes of MCF‐7, which is considered MSI stable (Ben‐David et al., 2018; Kleensang et al., 2016). In 2005, it was reported that during passaging of a couple of cell lines over a year, the profiles at several loci varied, with some alleles disappearing and then reappearing (Parson et al., 2005). Since then, these lines have been shown to be MSI unstable. Endometrial carcinomas are often mutated in the TP53 gene and show MSI (Berchuck & Boyd, 1995; Soleymani et al., 2026). In 1981, one of the first endometrial cancer cell lines, Ishikawa, was established from a well‐differentiated adenocarcinoma (Nishida, 2002). Three isolates of this cell line were shared widely by its originator (M. Nishida); the original culture (labeled “Original” herein) from 1981 to 1993, strain 3‐H‐4 from 1993 to 1995, and strain 3‐H‐12 from 1996 onwards. Nishida reported that these three strains differed considerably in phenotype. The Ishikawa cell line has been found to be MSI unstable and mutated in TP53 (Albitar et al., 2007; Devor et al., 2020; Korch et al., 2012) and as a result has many STR loci with multiple alleles. These findings have been subsequently confirmed in great detail (Devor et al., 2020). Supporting Information Tables S4 and S5 compare the profiles of 54 samples of Ishikawa strains, the ECC‐1 and EnCa101AE cell lines, and tissue samples (both authentic and false samples). From the labeling of the Ishikawa samples and their listings in Table 3 of Nishida (2002) and other resources, four STR profiles were chosen as references to query and identify the likely origins of the other 50 samples; unfortunately, many recipients of this cell lines did not carefully note which isolate was received and identified the samples only as “Ishikawa” or “ECC‐1.” These highly variable genotypes were analyzed using the three algorithms in the ICLAC Human STR Genotype Comparison Tool (.xlsm; Appendix 2 of the accompanying protocol article; Korch et al., submitted). Using the MvR algorithm, it was found that the purported independently isolated endometrial cell line ECC‐1, which was claimed to have been isolated from the xenograft tumor tissue EnCa101 (Satyaswaroop & Tabibzadeh, 1991), was actually a derivative of Ishikawa, probably from strain 3‐H‐12 (Korch et al., 2012). Clearly, it is not derived from any of the four EnCa101 samples (Supporting Data Table S4). Furthermore, as these 54 STR profiles were variable (see Table 1 in Korch et al., 2012, and Supporting Information Tables S4 and S5 in this protocol) in their degree of genotype similarity to the four query/reference profiles of the identified strains, each sample was assigned to whichever of the three isolate categories (Original, 3‐H‐12, or 3‐H‐4) its reference STR genotype best matched based on TNB and MvR scores (see the Original 3‐H‐12 3‐H‐4 worksheet in Supporting Data Table S5), but these assignments are not definitive. Moreover, four samples of ECC‐1 from different sources turned out to have been either mislabeled or cross‐contaminated and taken over by MCF‐7 cells, with one sample surviving as a mixture of MCF‐7 cells and the Ishikawa subline ECC‐1. An early passage of ECC‐1 (p44 from September 23, 1991), which was received in person from the originator, was expanded, and p49 was STR analyzed in 2012 and found to be MCF‐7 cells (Amanda Capes‐Davis, personal communication, March 1, 2026). This may explain why other samples of ECC‐1 were found to be MCF‐7. In 1990, it was reported that the ECC‐1 cells were derived from an endometrial adenocarcinoma (Satyaswaroop & Tabibzadeh, 1991; Tabibzadeh et al., 1990), which was subsequently maintained in mice as a tumor named as EnCa101AE (usually referred to as simply EnCa101). Four isolates of this tumor (three from 2009 and one from 1988) were received (in 2009 and 2012, respectively) from V. Craig Jordan, its sole known recipient/propagator, and STR analyzed. None were found to be related to any Ishikawa/ECC‐1 cell line isolates. The 1988 sample of EnCa101 (see DP0276‐01 and DP0276‐02, Supporting Information Table S4) could possibly, because of its age, be an actual sample of the EnCa101 tissue sample; unfortunately, the sample was not viable. This is clearly a major mishap for previous studies of endometrial cancer. Another misidentified cell line is the cell line Ishikawa (Heraklio) 02 ER with Cellosaurus identifier CVCL_6543. The STR genotype is from DepMap (The Cancer Dependency Map Project) at the Broad Institute of MIT and Harvard, where it has the identifier ACH‐000961 and is designated ISHIKAWAHERAKLIO02ER. However, its source is listed as the European Collection of Authenticated Cell Cultures (ECACC), with accession number 99040201. This catalog number is actually for Ishikawa cells and not Ishikawa (Heraklio) cells, which have the ECACC accession number 9803280. This explains why in Cellosaurus this cell line has a genotype similar to those of other Ishikawa‐derived lines. The actual Ishikawa (Heraklio) version, originally deposited in the ECACC in 1998, was found independently to be HEC‐50 cells by both me and the ECACC. The STR data presented in Cellosaurus do match those of Ishikawa 3‐H‐12, as does the TP53 mutation described in DepMap. It is the name of CVCL_6543 that is incorrect: it should be identified simply as Ishikawa, without mention of “Heraklio.” Table 9 summarizes the allelic frequency data from Supporting Information Tables S4 and S5, showing how variable the different STR profiles are at 15 STR loci; there are up to 15 different allelic genotypes at the 15 STR loci. From these frequency data, the most likely origins of the three Ishikawa strains (original, 3‐H‐4, 3‐H‐12, and ECC‐1) were deduced by selecting a genotype for each strain and comparing their TNB and MvR scores at the 15 loci (Supporting Information Table S5), and then comparing those with the genotype of the earliest sample of an Ishikawa strain (ECACC 99040201, deposited in the ECACC in 1999). They match closely, differing only at the D3S1358 locus by an extra allele that is one tetranucleotide repeat smaller than an allele in the “deduced” profile. Additionally, the likely genotypes of the four individual Ishikawa strains were deduced and are shown in Supporting Information Table S5. Table 9. | STR locus | Ishikawa variant genotype 1 | Ishikawa variant genotype 2 | Ishikawa variant genotype 3 | Ishikawa variant genotype 4 | Ishikawa variant genotype 5 | Ishikawa variant genotype 6 | Ishikawa variant genotype 7 | Ishikawa variant genotype 8 | Ishikawa variant genotype 9 | Ishikawa variant genotype 10 | Allele ratios in different STR genotypes (1:2:3:4:5:6:7:8:9 + 10–15) | Total no. profiles | Most common Ishikawa STR genotype | Genotype of Ishikawa “Original” sent to E. Gurpide | Genotype of Ishikawa 3‐H‐12 (#79) sent to BA Lessey | |---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---| | CSF1PO | 11,12 | 11,12, 13 | 11,13 | 28:8:1 | 37 | 11,12 | 11,12 | 11,12,13 | ||||||| | D2S1338 | 20 | 18,20 | 19,20 | 16,18,20 | 18:8:7:1 | 34 | 20 | 18,20 | 19,20 | |||||| | D3S1358 | 16,17 | 17,18 | 15,17 | 16,17,18 | 15,16,17 | 17 | 21:8:3:2:1:1 | 36 | 16,17 | 17,18 | 16,17 | |||| | D5S818 | 10,11 | 10,11, 12 | 36:3 | 39 | 10,11 | 10,11 | 10,11 | |||||||| | D7S820 | 9,10 | 8.3,11 | 8.3,11, 12 | 8.3,9,10, 11,12 | 30:7:1:1 | 39 | 9,10 | 8.3,11 | 9,10 | |||||| | D8S1179 | 12,16 | 12.13, 16 | 13,16 | 12,15,16 | 10,13,16 | 12,16,17 | 20:6:5:2:1:1 | 35 | 12,16 | 12,16 | 12,13,16 | |||| | D13S317 | 9,12 | 9,12,13 | 9,13 | 9,11,12 | 25:6:6:1 | 38 | 9,12 | 9,12 | 9,12,13 | |||||| | D16S539 | 9 | 8,9 | 9,10 | 36:2:1 | 39 | 9 | 9 | 9 | ||||||| | D18S51 #1 ‐ 10 + #11 ‐ 15 | 13,19 | 12,19 | 13,20 | 13,21,22 | 14,20,21 | 14,19, 20,21 | 12,19, 20 | 14,19, 20,21 | 14,20 | 12,13, 19,20 | 9:7:5:3:3: 2:2:2:1: 1:1:1:1:1 | 39 | 13,19 | 14,19,20, 21 | 12,19,20 | | 12,20 | 13,14, 18,19, 20 | 13,14, 19,21 | 14,19, 20 | 14,21 | ||||||||||| | D19S433 | 12.2,14 | 13.2,14 | 12.2,15 | 29:4:2 | 35 | 12.2,14 | 12.2,14 | 13.2,14 | ||||||| | D21S11 | 28 | 28,29 | 35:1 | 36 | 28 | 28 | 28 | |||||||| | FGA | 21 | 20,21 | 20 | 21,22 | 22:6:5:3 | 36 | 21 | 21 | 20,21 | |||||| | TH01 | 9,10 | 9 | 9,11 | 36:2:1 | 39 | 9,10 | 9,10 | 9,10 | ||||||| | TPOX | 8 | 38 | 38 | 8 | 8 | 8 | ||||||||| | vWA | 14,17 | 14,18 | 14,16,17,18 | 28:9:1 | 38 | 14,17 | 14,17 | 14,17 | The frequencies of STR alleles in 31 different STR genotypes of Ishikawa and the false cell line ECC‐1, taken from Supporting Information Tables S4 and S5. D18S51 showed the greatest number of variant combination of alleles. Based on the most frequent allele combinations, the most likely STR genotype was deduced and is compared to the genotypes of the most likely genotype of the “Original” Ishikawa cells sent to Erlio Gurpide at Mount Sinai Medical Center on September 10, 1984, from the originator's laboratory, and to the subline 3‐H‐12 (sample 79) sent to Bruce A. Lessey at the University of North Carolina on October 11, 2002. The genotypes for 3‐H‐12 samples were the most frequent among the 31 samples in Table S4. The Gurpide source of cell line samples have a unique 8.3 allele instead of a 9 allele at the D7S820 locus and a 13.2 allele at D19S433 instead of the more common allele of 12.2. Supporting Information Table S5 contains STR allele frequencies for all the samples and for the four different sublines. This case study of 54 cell line samples illustrates how to evaluate their relatedness despite the extensive variability in their STR genotypes due to MSI. It also emphasizes that cell lines can vary greatly, which can be a major concern for the reproducibility of published studies and work based on the earlier work. Clearly, slight differences in a cell line's STR genotype should be a “canary” in the cell line “minefield,” warning about the importance of characterizing a cell line's suitability more extensively than just by using STR genotyping. Masters versus Query (MvQ) Algorithm In a third situation, where two human cell lines have fused and formed a hybrid cell line, the resulting line will have STR markers from each of the fusion partners. Intraspecies human‐human somatic cell hybrids such as a HeLa + unknown Chinese cell line (Strong et al., 2014) and interspecies mouse‐human hybrids such as hybridomas (Chan et al., 2025; Jacobsen et al., 2006; Köhler & Milstein, 1975), arising by accidental or induced cell fusion, have been described since the late 1960s. The STR profiling data for the purported nasopharyngeal carcinoma cell (NPC) lines HNE1 (CVCL_0308), HONE1 (CVCL_8706), and CNE1 (CVCL_6888) are shown in Supporting Information Table S1 together with profiles of other HeLa‐related cell lines. Strong et al. showed that these lines carried the human papilloma virus HPV‐18 integrated in their genomes in patterns, along with single‐nucleotide variations (SNVs), identical to those observed in HeLa cells and concluded that these cells arose by somatic hybridization/fusion with HeLa cells (Strong et al., 2014). NPCs are associated with Epstein‐Barr virus (EBV), which can also be used to fuse cell lines and is found in >90% of the adult human population worldwide (Chesnokova et al., 2009). EBV has been used to create lymphoblastoid cell lines, such as the ML14 cell line (Korch et al., 2018; Saxton et al., 1978). This raises the question of whether laboratories studying NPC have inadvertently, through improper containment of cell cultures, given rise to fused hybrid cell lines, such as the above HeLa‐derived hybrids and hybrids with other NPC cell lines. Testing the Situation 3 hypothesis with the MvQ algorithm using HT‐29 genotypes In Situation 3, it is more difficult to determine cell line identity because the number of alleles available may be insufficient to yield a unique profile. For STR genotyping, it is generally recommended that, with a minimum of eight or nine STR loci, a score of 90% be used (Bady et al., 2012; Huang et al., 2017; Yu et al., 2015). Including additional STR loci improves the discriminatory power of STR analysis; currently, use of 13 or 16 loci is recommended (ASN‐0002, 2022). Huang et al. reported that a 21‐locus assay could distinguish between some cells lines that were not distinguishable when using eight or nine STR loci using the TNB algorithm (Huang et al., 2017). With these limitations in mind, we decided to test whether the MvQ algorithm could still be useful to identify cell lines with very low heterozygosity. The MvQ algorithm might be effective for comparing STR profiles of very low heterozygosity with STR profiles in a database such as Cellosaurus because the small number of alleles in the query profile limits the potential number of matches, as explained earlier. To test this hypothesis, the STR profile of the HT‐29 cell line was used unmodified (referred to as the “full‐heterozygous genotype” with 17 loci, of which are 13 heterozygous) and in a modified version from which six alleles (the larger, i.e., longer, ones) were removed from six of those heterozygous loci, leaving just 7 of the 17 STR loci heterozygous (the “half‐heterozygous genotype”). Two additional versions of the HT‐29 profile were created in which none of the 17 STR loci are heterozygous: Version 1 was modified to retain only the smaller (i.e., shorter) of the two alleles at each of the 13 originally heterozygous STR loci (“zero‐heterozygous with the smaller 1st‐allele genotype”), and Version 2 was modified to retain only the larger (i.e., longer) of the two alleles at each heterozygous STR locus (“zero‐heterozygous with the larger second‐allele genotype”). These four STR genotypes (Table 10) were then used in CLASTR to search for matches using all three algorithms (Tanabe, MvR, and MvQ), yielding 12 search results. Table 11 lists the 24 different HT‐29‐related cell lines, including the reference profile of the colon‐derived cell line HT‐29 (CVCL_0320) and 13 lines identified as known synonymous derivatives of HT‐29 plus 10 cell lines that have been found to be misidentified derivatives of HT‐29. Included in this list is the number of loci tested, the total number of alleles in the STR profiles, and the heterozygosity ratio. Table 10. | Sample name | CVCL_ 0320 1 | CVCL_ 0320 2 | CVCL_ 0320 3 | CVCL_ 0320 4 | CVCL_ 0320 5 | CVCL_ 0320 6 | |---|---|---|---|---|---|---| | Description | HT‐29 full‐Hz profile | HT‐29 half‐Hz profile | HT‐29 zero Hz, only 1st alleles | HT‐29 zero Hz, only 2nd alleles | HT‐29 zero Hz, only low‐frequency alleles | HT‐29 zero Hz, only high‐frequency alleles | | STR locus | |||||| | Amel | X | X | X | X | X | X | | CSF1PO | 11,12 | 11 | 11 | 12 | 11 | 12 | | D2S1338 | 19,23 | 19 | 19 | 23 | 13 | 19 | | D3S1358 | 15,17 | 15 | 15 | 17 | 17 | 15 | | D5S818 | 11,12 | 11 | 11 | 12 | 11 | 12 | | D7S820 | 10 | 10 | 10 | 10 | 10 | 10 | | D8S1179 | 10 | 10 | 10 | 10 | 10 | 10 | | D13S317 | 11,12 | 11 | 11 | 12 | 11 | 12 | | D16S539 | 11,12 | 11 | 11 | 12 | 12 | 11 | | D18S51 | 13 | 13 | 13 | 13 | 13 | 13 | | D19S433 | 14 | 14 | 14 | 14 | 14 | 14 | | D21S11 | 29,30 | 29,30 | 29 | 30 | 29 | 30 | | FGA | 20,22 | 20,22 | 20 | 22 | 20 | 22 | | Penta D | 11,13 | 11,13 | 11 | 13 | 13 | 11 | | Penta E | 14,16 | 14,16 | 14 | 16 | 16 | 14 | | TH01 | 6,9 | 6,9 | 6 | 9 | 9 | 6 | | TPOX | 8,9 | 8,9 | 8 | 9 | 9 | 8 | | vWA | 17,19 | 17,19 | 17 | 19 | 19 | 17 | 1, Full, with 13 of 17 loci heterozygous; 2, half‐heterozygous genotype with only 7 of 17 loci heterozygous; 3, zero‐heterozygous version 1 retaining only the smaller first allele of the original genotype; 4, zero‐heterozygous version 2, retaining only the larger second allele of the original genotype; 5, zero‐heterozygous version 3, using only the alleles with lower frequency of the two in human STR genotypes; 6, zero‐heterozygous version 4, using only the alleles with the higher frequency of the two in human STR genotypes. These were used to search CLASTR using the three algorithms to identify the scores of 24 known HT‐29 related cell lines listed in Table 11, as described in the text. Table 11. | No. | CVCL ID | Cell line name | STR loci | No. heterozygous STR loci | Total no. alleles | Heterozygosity ratio (Hz loci/total loci) | |---|---|---|---|---|---|---| | Different subclones of HT‐29 | |||||| | 1 | CVCL_0320 | HT‐29 reference | 17 | 13 | 30 | 0.76 | | 2 | CVCL_2011 | HT‐29/CX‐1 | 17 | 12 | 29 | 0.71 | | 3 | CVCL_A8P7 | HT29‐5F12 | 15 | 12 | 27 | 0.80 | | 4 | CVCL_A8P8 | HT29‐5F7 | 15 | 11 | 26 | 0.73 | | 5 | CVCL_2521 | HT29/219 | 13 | 10 | 23 | 0.77 | | 6 | CVCL_J256 | HT‐29‐Luc | 8 | 6 | 14 | 0.75 | | 7 | CVCL_A3CP | HT29‐5FU resistant 1 µM | 15 | 12 | 27 | 0.8 | | 8 | CVCL_A8P6 | HT29‐5FU resistant 10 µM | 15 | 8 | 23 | 0.53 | | 9 | CVCL_A8P9 | HT29‐5M12 | 15 | 8 | 23 | 0.53 | | 10 | CVCL_A8Q0 | HT29‐5M21 | 15 | 7 | 22 | 0.47 | | 11 | CVCL_F0KM | HT29‐MTX resistant 1 µM | 15 | 9 | 24 | 0.60 | | 12 | CVCL_F0KP | HT29‐MTX resistant 100 µM | 15 | 8 | 23 | 0.53 | | 13 | CVCL_F0KN | HT29‐MTX resistant 10 µM | 15 | 8 | 23 | 0.53 | | 14 | CVCL_G356 | HT‐29‐MTX‐E12 | 8 | 4 | 12 | 0.50 | | Misidentified/cross‐contaminants of colon‐derived cell line HT‐29 | |||||| | 15 | CVCL_0144 | ARO | 13 | 9 | 22 | 0.69 | | 16 | CVCL_6302 | KAT‐10 | 9 | 6 | 15 | 0.67 | | 17 | CVCL_6304 | KAT‐4 | 13 | 8 | 21 | 0.62 | | 18 | CVCL_0370 | KAT‐5 | 9 | 5 | 14 | 0.56 | | 19 | CVCL_L095 | KM20 | 9 | 6 | 15 | 0.67 | | 20 | CVCL_6310 | MRO | 9 | 6 | 15 | 0.67 | | 21 | CVCL_7204 | MV‐522 | 8 | 7 | 15 | 0.88 | | 22 | CVCL_0E37 | MV522/MDR1 | 8 | 5 | 13 | 0.63 | | 23 | CVCL_0E38 | MV522/MRP | 8 | 5 | 13 | 0.63 | | 24 | CVCL_2760 | WiDr | 17 | 13 | 30 | 0.76 | Data included are number of STR loci tested, number of heterozygous loci, total number of alleles in the STR profiles, and heterozygosity ratio. The STR genotypes are from Cellosaurus and are shown in Table S6. Table 12 summarizes the results obtained by searching for the 24 HT‐29‐derived cell lines using the three algorithms and the four versions of the HT‐29 STR genotype as the query profiles (see Supporting Information Table S6 for all CLASTR search results). Table 12 includes the (1) ratio of heterozygosity of the four similarity scores calculated by the CLASTR tool, (2) the range of scores, (3) the number of STR profiles of unrelated cell lines that had scores which were within the range of scores spanned by the different versions of HT‐29, (4) the standard deviation of the similarity scores obtained using the three different algorithms on the unmodified genotype of HT‐29 in Cellosaurus, (5) the modified genotype with approximately half the original heterozygosity and the two versions of the HT‐29 genotype with no heterozygous loci, (6) Pearson's correlation coefficients of the sample heterozygosities versus match scores, and (7) the random match probabilities (RMPs) of the different query profiles based on the NIST allelic frequencies (Butler et al., 2025; Gettings et al., 2018; Ruitberg et al., 2001; Steffen et al., 2017). Table 12. | HT‐29 STR query genotype | No. STR loci | No. hetero‐zygous loci | Hetero‐zygosity ratio (Hz loci/total loci) | No. HT‐29‐related cell lines identified | Analysis algorithm | Range of similarity % scores including all known HT‐29‐related cell lines | No. of unrelated cell lines in HT‐29 similarity range | Mean similarity scores (%) | Standard deviation (%) | Sample Hz vs. match score Pearson's correlation coefficient (ρ) a | Random match probability NIST allelic frequencies | |---|---|---|---|---|---|---|---|---|---|---|---| | Full heterozygous | 17 | 13 | 0.765 | 24 | TNB | 86.7%‐100% | 0 | 92.4% | 4.6% | +0.575 | 2.13 × 10−22 | | MvR | 86.7%‐100% | 1 | 96.1% | 5.2% | −0.080 | |||||| | MvQ | 80.0%‐100% | 0 | 89.2% | 6.1% | +0.851 | |||||| | Half heterozygous | 17 | 7 | 0.412 | 24 | TNB | 75.0%‐90.6% | 4 | 83.6% | 5.6% | +0.300 | 9.22 × 10−19 | | MvR | 69.2%‐86.4% | 12 | 77.2% | 5.1% | −0.212 | |||||| | MvQ | 81.8%‐100% | 4 | 90.4% | 6.4% | +0.689 | |||||| | Zero heterozygous with 1st (smaller) alleles | 17 | 0 | 0 | 24 | TNB | 64.7%‐81.1% | 30 | 72.4% | 4.9% | −0.485 | 1.5289 × 10−12 | | MvR | 52.2%‐68.2% | 49 | 58.2% | 4.9% | −0.655 | |||||| | MvQ | 84.6%‐100% | 65 | 96.0% | 5.5% | +0.064 | |||||| | Zero heterozygous with 2nd (larger) alleles | 17 | 0 | 0 | 24 | TNB | 50.0%‐72.74% | 140 | 63.1% | 6.4% | +0.714 | 1.3202 × 10−13 | | MvR | 41.7%‐57.1% | 98 | 50.5% | 4.5% | + 0.559 | |||||| | MvQ | 62.5%‐100% | 312 | 84.7% | 10.6% | +0.834 | |||||| | Zero heterozygous with low‐frequency alleles | 17 | 0 | 0 | 24 | MvQ | 75.0%‐100% | 92 | 88.4% | 8.3% | +0.661 | 3.782 × 10−14 | | Zero heterozygous with high‐frequency alleles | 17 | 0 | 0 | 24 | MvQ | 80.0%‐100% | 91 | 92.0% | 6.0% | +0.708 | 5.458 × 10−12 | Six types of genotypes—full heterozygous, half heterozygous, zero heterozygosity with only 1st allele, zero heterozygosity with only 2nd allele, xero heterozygosity with only the lower‐frequency allele, and zero heterozygosity with only the higher‐frequency allele—were used to find the 24 cell line variants of HT‐29 using the three search algorithms (TNB, Tanabe; MvQ, Masters versus Query; MvR, Masters versus Reference). Data included: Numbers of STR loci, heterozygous loci, heterozygosity ratios, HT‐29 related reference cell line genotypes, search algorithms, range of similarity scores, unrelated cell lines in range of HT‐29 related genotypes, mean similarity scores, standard deviation of scores, Pearson's correlation of heterozygosity ratios to match scores; and random match probability of NIST allelic frequencies. Ranges of positive and negative correlation coefficients: very weak, no correlation (white), ρ < ±0.2; weak (white), ρ = ±0.2 to < ±0.4; moderate (yellow), ρ = ±0.4 to < ±0.6; strong (orange), ρ = ±0.6 to < ±0.8; very strong (red), ρ = ±0.8 to ±1.0. From the results in Table 12, the following can be noted. Searching for matches using the full HT‐29 (CVCL_0320) reference STR genotype, all three algorithms (TNB, MvR, and MvQ) identified the 24 cell lines with scores between 80.0% and 100% identity and standard deviations between 4.6% and 6.1%. The TNB and MvQ algorithms did not identify any unrelated cell lines with match scores within the of ranges of scores for HT‐29 cell lines, and the MvR algorithm found only one unrelated cell line in the range for all HT‐29 samples. Reducing the total heterozygosity from 13 to 7 heterozygous loci (ratio decrease from 0.77 to 0.41) in the “half‐heterozygosity” query profile lowered the mean match scores calculated by the three algorithms, expanded the range of scores using the TNB and MvR algorithms for HT‐29 related cell lines, and increased the number of unrelated cell line whose profiles fell in the range for the HT‐29 cell lines (i.e., from 0 or 1 to 4 or 12 unrelated cell lines). However, the reduction of heterozygosity did not lower the MvQ range of scores and only slightly increased the number of unrelated cell lines with scores in the range of scores for HT‐29 cell lines (from 0 to 4). These results reflect the slightly decreased specificities of these scoring algorithms due to the reduction of heterozygosity. With the two “zero‐heterozygosity” profiles, in which all the STR loci were hemizygous (had only one STR allele per locus), the range of TNB and MvR scores for HT‐29 cell lines dropped even further, with maxima of <81.1% and minima as low as 41.7%. Moreover, the number of unrelated cell lines in these ranges increased severalfold (from 4.1‐ to as much as 35‐fold). In contrast, the MvQ scores ranged from 100% down to either 84.6% or 62.5%, depending on which alleles had been deleted. Again, as for the other algorithms, the number of unrelated cell lines with scores in the range of HT‐29 scores increased significantly (from 16‐ to 78‐fold). This increase is reflected in a doubling of the standard deviation value for the MvQ match scores of the HT‐29 samples. These results are displayed in bee‐swarm plots (Fig. 2), which illustrate how the TNB and MvR mean scores decreased with the reduction of heterozygosity of the HT‐29 query sample. This overall reduction in the mean scores did not occur with the MvQ results, even though the range of scores expanded greatly and the number of unrelated cell lines increased quite radically (Table 12). The inclusion of unrelated cell lines in the score ranges for the samples with reduced or no heterozygosity is due to the reduced specificity/discrimination that occurs with fewer alleles in the query profile and the increasing potential of the remaining STR alleles also being present in other, unrelated cell lines. On comparing the results for the two zero‐heterozygosity profiles, it is notable that better specificity is obtained (4.8‐fold fewer unrelated cell lines in the HT‐29 range of samples) when searching for matches to the sample containing only the first (smaller) allele present in the full HT‐29 genotype than when the query genotype contains the second (larger) allele of the full genotype. To look for possible causes of this phenomenon, the effect of heterozygosity in the reference samples on the specificity of the search algorithms was examined. Nine of the 24 STR profiles of HT‐29 cell lines had only eight or nine STR loci, while the remaining 15 had data for 13, 15, or 17 STR loci (Table 11). Figure 3 compares the match scores of these two zero‐heterozygosity query genotypes for the different HT‐29 related cell lines with their ratios of heterozygosity. It can be seen from the Pearson's correlation coefficients of the match scores to the two versions of the query genotypes that the zero‐heterozygosity genotype with the larger (second) alleles increased with increasing heterozygosity (r = +0.834), in contrast to what was observed using the zero‐heterozygosity genotype with only the smaller first alleles (r = +0.064). This difference is reflected in the greater number of unrelated cell lines in the range of HT‐29‐related cell lines using the genotype with the larger second alleles and the doubling of standard deviation calculated for these results from 5.5% to 10.6% (Table 12). When the TNB match scores of the HT‐29 family of genotypes obtained using the different HT‐29 query genotypes are compared, the zero‐heterozygosity first‐allele query genotype shows a negative Pearson's correlation of r = −0.485, whereas the other three TNB analyses yield scores between r = + 0.300 and + 0.714 (Table 12, Fig. 4). Negative correlations are seen for the MvR scores using the full‐heterozygosity, half‐heterozygosity, and first‐allele zero‐heterozygosity query genotypes, but not for the second‐allele zero‐heterozygosity query genotype (Table 12). As noted, the two HT‐29 zero‐heterozygosity genotype versions produced significantly different results. The zero‐heterozygosity profile with only the smaller first alleles had a mean score of 96.0% and a standard deviation of 5.5%, whereas the profile with only the larger second alleles had a mean score of 84.7% and standard deviation of 10.6%. This difference was reflected in the distribution of the HT‐29 family of genotypes obtained using the zero‐heterozygosity genotype, as only those searched using first alleles as the query were clustered more compactly in the bee‐swarm plot than those searched using only second alleles as query (Fig. 2). The cause of this significant difference was investigated by examining whether the frequencies of the STR alleles in human populations are correlated to this phenomenon. The RMP of the two genotypes was estimated using the STRidER human population allele probability calculator (https://strider.online/query) and the allele frequencies in the NIST STRbase 2.0 database (https://niststrbase.gov). As shown in Table 13, the RMP for these two genotypes showed an average 10‐fold difference, with the genotype with the zero‐heterozygosity first alleles being the larger by 8.57‐ or 11.6‐fold. Table 13. | Locus | First allele (smaller) | NIST allele freq. | 2nd allele (larger) | NIST allele freq. | Low‐freq. alleles | NIST allele frequency | High‐frequency allele | NIST allele frequency | T241 allele | NIST allele frequency | |---|---|---|---|---|---|---|---|---|---|---| | CSF1PO | 11 | 0.2736 | 12 | 0.3446 | 11 | 0.2736 | 12 | 0.3446 | 12 | 0.2953 | | D2S1338 | 19 | 0.1486 | 23 | 0.1182 | 23 | 0.1182 | 19 | 0.1486 | 20 | 0.1038 | | D3S1358 | 15 | 0.3045 | 17 | 0.2042 | 17 | 0.2042 | 15 | 0.3045 | 15 | 03085 | | D5S818 | 11 | 0.3156 | 12 | 0.3553 | 11 | 0.3156 | 12 | 0.3553 | 13 | 0.2237 | | D7S820 | 10 | 0.2949 | 10 | 0.2949 | 10 | 0.2949 | 10 | 0.2949 | 11 | 0.2302 | | D8S1179 | 10 | 0.0787 | 10 | 0.0787 | 10 | 0.0787 | 10 | 0.0787 | 12 | 0.1301 | | D13S317 | 11 | 0.2905 | 12 | 0.3050 | 11 | 0.2905 | 12 | 0.3050 | 12 | 0.4181 | | D16S539 | 11 | 0.2915 | 12 | 0.2568 | 12 | 0.2568 | 11 | 0.2915 | 9 | 0.1827 | | D18S51 | 13 | 0.1371 | 13 | 0.1371 | 13 | 0.1371 | 13 | 0.1371 | 19 | 0.0994 | | D19S433 | 14 | 0.3041 | 14 | 0.3041 | 14 | 0.3041 | 14 | 0.3041 | 14 | 0.2105 | | D21S11 | 29 | 0.2042 | 30 | 0.2476 | 29 | 0.2042 | 30 | 0.2476 | 30 | 0.2476 | | FGA | 20 | 0.0883 | 22 | 0.1974 | 20 | 0.0883 | 22 | 0.1974 | 21 | 0.1228 | | Penta D | 11 | 0.1556 | 13 | 0.1386 | 13 | 0.1386 | 11 | 0.1556 | 9.3 | 0.0965 | | Penta E | 14 | 0.0656 | 16 | 0.0507 | 16 | 0.0507 | 14 | 0.0656 | 8 | 0.3680 | | TH01 | 6 | 0.1959 | 9 | 0.1689 | 9 | 0.1689 | 6 | 0.1959 | 15 | 0.1915 | | TPOX | 8 | 0.4662 | 9 | 0.1377 | 9 | 0.1377 | 8 | 0.4662 | 12 | 0.2953 | | vWA | 17 | 0.2621 | 19 | 0.0787 | 19 | 0.0787 | 17 | 0.2621 | 20 | 0.1038 | | Average frequency | – | 0.2280 | – | 0.2011 | – | 0.1847 | – | 0.2444 | – | 0.2153 | | Random match probability (RMP) methods | RMP of HT‐29 genotype with only 1st alleles | RMP of HT‐29 genotype with only 2nd alleles | RMP of HT‐29 genotype with only low‐frequency alleles | RMP of HT‐29 genotype with only high‐frequency alleles | RMP of T241 genotype | ||||| | Strider hemiallelic a | 1.407 × 10−12 | 1.641 × 10−13 | 8.270 × 10−14 | 2.799 × 10−12 | 1.638 × 10−11 | ||||| | NIST hemiallelic | 1.410 × 10−12 | 1.408 × 10−13 | 3.489 × 10−14 | 5.692 × 10−12 | 2.094 × 10−11 | The frequencies of the alleles in human populations at 17 STR loci of the in silico‐modified genotypes of HT‐29 containing either the smaller 1st alleles, the larger 2nd alleles, the lower‐frequency alleles, or the higher‐frequency alleles. From these frequencies, the random match probabilities of each hemizygous genotype was determined using data from NIST (https://strbase.nist.gov/Information/NIST_Population_Data#1036SB). The probabilities in the second part of the table were calculated either using the online query tool of STRidER (https://strider.online/query) or by manually multiplying together the individual NIST allele frequencies. The calculations are based on two different population, which explains the slight differences between the NIST and STRidER totals. The genotype of the hemizygous cell line T241 (CVCL_M975) was included; its significance is discussed in the text. HT‐29 cells were derived from a Caucasian individual with 63% Northern European and 26% Southern European ancestry. Allele frequency data for all loci are from NIST (https://strbase.nist.gov/Information/NIST_Population_Data#1036SB). Frequencies from STRidER calculated using the tool at https://strider.online/query. Random match probabilities from STRidER were calculated using https://strider.online/query. A second test of STR low‐heterozygosity query genotypes was performed using a different selection of HT‐29 alleles. Two alternative HT‐29 zero‐heterozygosity allelic mixtures were created in silico, one consisting of the HT‐29 alleles with the lower allele frequencies and the other consisting of the HT‐29 alleles with higher frequencies. The HT‐29 query genotype with the higher‐frequency alleles yielded a much greater RMP than the one with lower‐frequency alleles (33.8‐ vs. 163‐fold). These two HT‐29 zero‐heterozygosity query genotypes were used to screen for cell line matches using CLASTR. The two zero‐heterozygosity genotypes with the larger RMPs (zero‐Hz first and zero‐Hz high‐frequency alleles in Table 13) identify HT‐29 sublines more specifically, with greater MvQ match scores and fewer unrelated cell lines in the range of MvQ scores for HT‐29 sublines (Table 12). Testing MvQ for actual hemizygous STR genotypes From these in silico experiments, it can be seen that low levels of heterozygosity in query samples can reduce the TNB and MvR scores to below the usually recommended 80% or 90% cutoff levels, but the MvQ algorithm can identify authentic matches with many scores greater than the cutoff levels found using TNB and MvR. However, low heterozygosity in the query sample has the drawback of increasing the number unrelated cell lines identified by the MvQ search algorithm. To test these observations further, we searched for cell lines with low or no heterozygosity. The thyroid cell line T241 (CVCL_M975) is monoallelic at all 15 STR loci. On searching for potential matches in the Cellosaurus CLASTR database, the TNB and MvR algorithms identified only the T241 cell line with scores of 100%; all other cell line genotypes gave scores of <75%. The MvQ algorithm identified T241 with a score of 100%, plus two unrelated cell lines with scores of 100%, one at 90%, and 26 additional unrelated cell lines with scores between 80% and 90%; but all of these unrelated cell lines had genotypes based on only 8 or 10 STR loci (data not shown). All of the unrelated cell lines had multiple diallelic loci, some of which are shown in red font to indicate discordancy between the profiles. The T241 genotype had an average allele frequency of 0.2153 and an RMP of 2.094 × 10−11 (Table 13). From these analyses, greater specificity is associated with average allele frequencies >0.21 and RMP estimates of ≥1.4 × 10−12 (see Table 13). This loss or lack of specificity is due to the limited number of alleles, which may be improved by increasing the number of alleles in the query sample(s) by analyzing additional STR loci, as suggested earlier (Huang et al., 2017). Examining the correlations of MvQ percent match scores and degrees of heterozygosity for the different HT‐29 samples using different algorithms (Figs. 3 and 5) indicated that the point of intersection of the correlations is at ∼0.8 heterozygosity, which would yield scores of ≈96% or greater. This level of heterozygosity suggests that ∼30 or more STR alleles are needed for optimal identification of a cell lines: that is, 30 STR hemiallelic loci or 13 of 17 STR loci being heterozygous. Figure 6 compares the correlation of TNB scores for full heterozygosity and MvQ scores for full heterozygosity, half heterozygosity, and zero heterozygosity with low‐frequency alleles versus the number of alleles of the family of HT‐29 cell genotypes with data for 13 or more STR loci. The Pearson's linear correlations converge at 100% match scores with 30 alleles. These results support the hypothesis that ideally STR genotyping should analyze a sufficient number of STR loci that a total of 30 alleles are included in profiles. The newer STR kits, which test over 20 STR loci, offer the advantage that samples with low heterozygosity can be more reliably authenticated. A complementary approach to distinguish between related cell lines from unrelated cells that have similarly high match scores between the query and the reference samples is to test the query sample for SNPs known to be in the various reference samples by PCR amplification and DNA sequencing or even through low‐read next‐generation DNA sequencing. An earlier report (Liang‐Chu et al., 2015) reviewed nine assays using between 21 and 92 (average 46) single‐nucleotide polymorphisms (SNPs) for the authentication of human cell lines and tested the use of 24, 48, and 94 SNPtrace assays. Although they found that having 24 and 48 SNPs was useful, as also described by others (Castro et al., 2013; Lee et al., 2005; Weiskirchen, 2026), they recommended the use of 96 SNPs, of which 46 are polymorphic alleles. The combination of these results with ours based on using STR alleles reinforces the finding that 30 STR alleles per genotype is a useful goal. CONFUSION AND QUESTIONABLE RESULTS IN THE BIOMEDICAL RESEARCH LITERATURE As reviewed above and in the accompanying protocol article (Korch et al., submitted), trying to identify and use authentic cell lines can be a daunting task when encountering a plethora of cell lines that are all seemingly ideal models for use by both the novice and the trained researcher. Imposter cell lines identified in the 1960s continue to be used by laboratories worldwide as models for their false identities (see examples in Fig. 1, Table 14, and Gorphe, 2019, Horbach & Halffman, 2017, Korch & Capes‐Davis, 2021, and Korch & Varella‐Garcia, 2018). Table 14. | Claimed false tissue origin | Imposter cell line names | True cell line and tissue origin | Publications June 2026 | |---|---|---|---| | Prostate | EB‐33 (CVCL_8344) H‐494 (CVCL_7669) HPC‐36M (CVCL_8446) KP‐P1 (CVCL_D283) MA‐160 (CVCL_8261) PC‐93 (CVCL_4888) | HeLa cervical adenocarcinoma | PubMed = 18 PMC = 85 | | Nasopharyngeal cancer | HEp‐2 (CVCL_1906) | HeLa cervical adenocarcinoma | Highwire (1955‐2014) = 8497 PubMed = 2334 PMC = 12,407 | | Normal embryonic intestinal tissue | Intestine 407 (CVCL_1907) | HeLa cervical adenocarcinoma | PubMed = 581 Highwire (1955‐2014) = 1397 PMC = 1968 | | Oral epidermal cancer | KB (CVCL_0372) KB‐3‐1 (CVCL_2088) + 321 descriptions and alternative names | HeLa cervical adenocarcinoma | PubMed = 959 PMC = 2639 | | Normal hepatic cells | Chang liver (CVCL_0238) | HeLa cervical adenocarcinoma | PubMed = 264 PMC = 1081 | | Breast cancer | MDA‐MB‐435 (CVCL_0417) MDA‐MB‐435S (CVCL_0622) Plus 21 sublines | M14 and sublines from a cutaneous melanoma from a male donor at a metastatic site | PubMed = 1099 PMC = 6141 | | Uveal melanoma | MUM‐2C (CVCL_3448) OCM‐1 (CVCL_6934) OCM‐1A (CVCL_6935) | M14 and sublines from a cutaneous melanoma from a male donor at a metastatic site | PubMed = 113 PMC = 333 | | Thyroid cancer | BHP 5‐16 (CVCL_6284) BHP 14‐9 (CVCL_6279) BHP 15‐3 (CVCL_6280) BHP 17‐10 (CVCL_6281) NPA (CVCL_0467) & 7 variants | M14 and sublines from a cutaneous melanoma from a male donor at a metastatic site | PubMed = 124 PMC = 605 | | Endometrial carcinoma | hTERT‐EEC‐B37 (CVCL_X007) IHEEC (immortalized human endometrial epithelial cell; not in Cellosaurus, but same as CVCL_X007) | MCF‐7 (CVCL_0031) invasive breast carcinoma | PubMed = 3 PMC = 22 | | Endometrial carcinoma | ECC‐1 (CVCL_7260) | MCF‐7 (CVCL_0031) OR Ishikawa (CVCL_2529) | PubMed = 121 PMC = 437 | | Ovarian adenocarcinoma | BG‐1 (AMC DP0126‐04) | Most are MCF‐7 (CVCL_0031) invasive breast carcinoma | PubMed = 146 PMC = 555 | Summarized are examples of misidentified cell lines and the prevalence of articles using these false/imposter cell lines under their false identities. The cell lines often have multiple names and descriptions of their tissue origins. Also, in many cases it is difficult to find articles using some cell lines because their names are not unique or standard (e.g., include KB vs. kb for kilobase and other variants), which requires complex search algorithms in PubMed and PubMed Central. The numbers of articles are estimates of the prevalence of articles referring to cell lines incorrectly. Parson et al. (2005) illustrated that the STR profiles of cell line cultures can vary over time, with alleles disappearing and reappearing. This is partly explained by the subsequent finding that some of the cell lines they studied were MSI unstable, which can give rise to STR alleles that can increase or decrease in length by one tetranucleotide repeat unit over time. However, even MSI‐stable cells can vary in the gain or loss of alleles, as exemplified in Tables 1, 2, 3, 4, 8, and 9 and Supporting Information Tables S1‐S5. Cell lines are often renamed, both inadvertently and purposefully, which adds confusion to the biomedical literature. As shown in Table 2, the family of eight colon cancer cell lines (DLD‐1, HCT‐8, HCT‐15, HRT‐18, C146, C170, MIP‐101, and T‐47D‐Luc) are related by their STR genotypes, showing that they all originated from the same donor, but have different names. A more serious and difficult aspect of confusion over cell line identities is exemplified by the report from Han et al. (2012). The authors received one of the hTERT‐EEC clones (see Table 3 and Supporting Information Table S2) from the Hombach‐Klonisch and Klonisch laboratory (Hombach‐Klonisch et al., 2005), renamed it as an immortalized human endometrial epithelial cell (IHEEC) line, and used it in a co‐culture experiment with an immortalized human endometrial stromal (IHESC) cell line to show that it could be grown in mice if co‐cultured with an epithelial cell line, that is, IHEEC/hTERT‐EEC. They stated that this was done to model endometriosis by emulating the communication necessary between stromal and epithelial cells. However, when informed that hTERT‐EEC cells were actually MCF‐7, they responded that it did not matter for the validity of their article. However, this was not a valid test because MCF‐7 is a pleural effusion metastasis‐derived breast cancer cell line that behaves differently from endometrial cell lines, and furthermore it has been used extensively to establish xenografts in SCID mice by itself. A PubMed search for “(MCF‐7 OR MCF7) AND SCID” yielded 675 publications, whereas substituting “xenograft” for “SCID” found 5011 publications, as of January 2026. Clearly, this was not a valid test for this stromal‐epithelial cell interaction. The authors did not authenticate their IHEEC cell line, nor did they test which cell line gave rise to the cells that grew in the SCID mouse. Was it simply MCF‐7 cells? This renaming and claiming of other cell line properties definitely pollutes the cell line literature and promotes the waste of resources. SUMMARY OF BEST LABORATORY STEPS FOR ANALYTICAL CELL LINE AND TISSUE AUTHENTICATION Fortunately, Cellosaurus, ICLAC, and several other reviews (Gorphe, 2019; Horbach & Halffman, 2017; Korch & Capes‐Davis, 2021; Korch et al., 2012; Neimark, 2014, 2015; Schweppe et al., 2008; Vaughan et al., 2017) have revealed the relatedness of many false cell lines and tissue samples. Still, a great deal of the biomedical research literature has been polluted with extensive questionable research based on imposter cell lines. The purpose of the laboratory practices presented here and the accompanying protocol article (Korch et al., submitted) is to provide an understanding of the analytical process necessary to authenticate cell samples, using many examples to illustrate the types of data that a researcher will encounter and need to interpret. From the extensive discussions above, it is obvious that STR genotyping results are influenced by many factors. Therefore, careful handling of cell lines and understanding of STR data are critical for obtaining valid research results. Outlined below and in Supporting Information Appendix 1 are steps that biomedical research institutions should establish as standard laboratory protocols in which all lab personnel are trained. Initial Cell Line Pre‐Ordering Steps Do the following before ordering any cell line for a project. - 1. Check Cellosaurus for any information known about the cell line's provenance, its authenticity, and whether it has been STR genotyped. - 2. If it has been genotyped, use all three algorithms in CLASTR to check its identity. Are there any potential matches? As the Cellosaurus database grows, new profiles are added to it several times per year, so a new search may identify a new match. - 3. If the cell line is not described in Cellosaurus, check with the originator for what is known about it, such as (i) whether it was obtained from a reliable source, (ii) whether it has been STR genotyped, (iii) whether the number of passages it has undergone is known, and (iv) whether it has been regularly STR authenticated. Also, ask whether there are there any tissue samples of the original donor from which DNA is available for STR genotyping. - 4. Obtain the STR genotyping data (ideally the actual electropherograms) and use CLASTR to see whether there are any matches to them, using all three algorithms (TNB, MvQ, and MvR). - 5. Do the electropherograms show any evidence of minor peaks indicating the presence of contaminating cell lines? Contaminating cell lines can take over a culture. Establishment of Cell Line Information Records and STR Genotyping Database Set up a database containing the history for each cell line that is currently being used in the laboratory, which should consist of the following. - 1. Create an individual file describing each cell line that can be edited to allow for the addition of new information; - 2. Using Worksheet 3 of Appendix 2 (ICLAC Human STR Genotype Macro Comparison Tool, .xlsm) from the accompanying protocol article (Korch et al., submitted), start creating a set of STR genotypes as the lab accumulates STR data for all cell lines in use; - 3. Keep records of all treatments of the cell line in the above editable file; and - 4. Keep a record of each passage with date, approximate growth phase (log vs. stationary), observations of any unusual cell morphology, etc., passage dilution, and whether a sample of residual cell suspension centrifuged and stored at −20°C is available for STR genotyping and downstream troubleshooting of cell line identity (see next step below). Initial Culture of Cell Line or Tissue Sample in Preparation for STR Genotyping Do the following upon receiving the cell line (in a frozen ampule or culture flask). - 1. Grow up an aliquot of the cells in the appropriate medium in a separate “Quarantine” incubator that is used solely for growing new cultures before their use in the laboratory. An ideal project insurance policy is, if possible, to preserve/retain a sample of the original tissue/donor sample (tumor, xenograft, pleural effusion, blood, etc.) from which the working cell sample was derived. This will allow confirmation that the STR genotype of the cell line corresponds to that of a donor sample. Also, as genetic methods evolve, newer questions may arise that require original tissue samples, including confirmation of the original identity. - 2. Dedicate a separate fresh bottle of culture medium for each cell line. Do not use it for any other cultures. - 3. For the cell culture log, note any unusual features of the growing cells, such as cell morphology, indications of multiple phenotypes or morphologies, growth rates, special medium requirements, and manner of growth (attached or suspended), and note these features in a digital database file for each cell line. - 4. On performing the first passage—for example, a 1:10 dilution—save the remaining cell suspension for extraction of DNA. A second simple laboratory insurance policy is to save this unused cell suspension by centrifuging down the cells; removing most of the supernatant, leaving about 1 ml of residual medium; resuspending the cell pellet; and transferring it to a 1.7‐ to 2.2‐ml microcentrifuge tube. Label the tube with the cell line name, passage number, and date, and then store the tube in a box designated for that specific set of samples at −20°C (cell viability is not necessary for these samples). Printed, instead of handwritten, labels are best to avoid ambiguity in handwritten letters or numbers (e.g., S vs. 5, l [L] vs. 1 [one] or 0 vs. O). DNA can later be extracted from this sample of cells for STR genotyping if you ever need to determine at what point in a series of experiments using this cell line it was no longer correctly identified. - 5. Keep the passaged cells in the Quarantine incubator. - 6. Check an aliquot of the cell suspension for mycoplasma using a PCR kit that tests for a broad spectrum of mycoplasma. - 7. Extract the DNA from the unused cell suspension sample using a kit, such as the Zymo Research Quick‐DNA Miniprep D3024. Estimate the DNA concentration of the sample by running a 5‐µl aliquot of the extracted DNA on a 0.75% agarose gel (with DNA stain, e.g., ethidium bromide, SYBR Safe, GelRed, or GelGreen) together with a DNA Mass Ladder (e.g., BioLine HyperLadder 1 kb, available at https://www.bioline.com/hyperladder), which is a size ladder with known amounts of DNA per DNA fragment/band. Photograph the gel with an exposure that shows the intensities of the bands to be proportional to the amounts in each band, that is, such that the band intensities range proportionally from faint gray to strong white against a black background. By comparing the sample bands to those in the mass ladder lane, one can estimate the DNA concentration of the sample. This provides semiquantitative estimates of sample DNA concentrations and allows one to see whether the sample contains detectable amounts of RNA that would inflate the OD260 measurements used to estimate DNA concentration. Protein, very large DNA, and unlysed cells can cause DNA staining to be trapped in the gel well. RNA, protein, and cell debris can inflate any OD260 estimates of DNA concentrations, especially when measuring small microliter volumes by absorption or light scattering. For these reasons, nanometric OD260 measurements of DNA concentrations can be unreliable. However, if a nanometric OD260 instrument must be used, measure three separate aliquots of a sample and ascertain whether the values are within the instrumental detection limits. An accurate estimate of DNA concentrations is essential for reliable STR genotyping results. Use a standard of known concentration to confirm that the values determined are reliable. - 8. Depending on the source of the cell line and its planned uses, cell samples may be cross‐contaminated with cells from nonhuman mammalian species such as mouse or another species being used in the laboratory. This can occur when using xenografts to establish cell lines or in studies of tumorigenicity of a cell sample. A PCR assay method has been published that can screen for the presence of cells commonly used in tissue culture labs (Cooper et al., 2007). One of the assays checks for two species‐specific DNA sequences in mitochondrial genes, cytochrome C oxidase subunit I and cytochrome B. Some genotyping facilities offer this analysis as a service for samples submitted for STR genotyping. Alternatively, this simple PCR‐based assay can be set up in most laboratories and is especially useful for those working with xenografts. But because this tests for specific mitochondrial DNA segments, it can only detect intact cells and not genomic DNA that would occur in interspecies cell fusions. - 9. Cell line fusions can occur not only between two human cell lines, as described above, but also between human and mouse cell lines (Chan et al., 2025; Ephrussi & Weiss, 1965; Harris & Watkins, 1965; Jacobsen et al., 2006; Nims et al., 2018; Saxon et al., 1985; Weiss & Green, 1967). The above PCR assay for testing interspecies cross‐contamination checks only for mitochondrial DNA, but a check for genomic DNA would be more helpful for detecting interspecies cell fusion. A PCR assay for a genomic DNA sequence specific for the organisms of concern would be useful. One such option to identify potential fusions with mouse cells is the STR authentication assay for mouse STR loci (Almeida et al., 2014, 2019; Almeida & Korch, 2023). The current version of this assay identifies alleles at 18 mouse genomic STRs and two human genomic STRs (as a check for human cell cross‐contamination). This mouse‐specific STR assay is offered by a few service facilities and may be an option to consider when choosing a facility. - 10. Identify the service laboratory that will perform the STR genotyping assay. Check that it uses all of the following. - i. An STR kit that assays a minimum of 17 STR loci. As discussed above, kits that test 20 or more STR loci are better, especially for samples with low degrees of heterozygosity. - ii. Confirm that you will receive from the service facility both the STR electropherogram of your sample and those of the positive DNA control (PDC) and the no‐DNA control (NDC). The NDC peaks should generally be <100 relative fluorescence units (RFU). Before using the service, ask to see examples of the STR electropherograms and other data that you would receive. - i. Are the peaks in the electropherograms adjusted so that at least one peak in each color pane fills the panes? It not, this is probably due to incorrect settings in the allele‐calling software, and this could make it difficult to detect smaller amounts of DNA from a contaminating human cell sample. Check that the peaks are labeled both with their size (“sz”; in bp length) and allele call. Are the sample peaks well above the signal levels of the NDC background levels? Generally, you want the allele peaks to be between 1000 and 20,000 RFU in intensity. The values will vary with locus, allele, STR kit, capillary instrument used, DNA concentration, amount of sample loaded, and sample loading buffer. Check what volumes and concentrations of DNA are required by the service facility. - 11. Once satisfied with the quality of the facility's service, submit your DNA sample as required by the chosen facility. Quality Assessment of STR Data Upon receipt of the electropherograms for your DNA sample(s) and those of the control samples (PDC and NDC), evaluate the quality of the STR data according to the following criteria. - 1. In the NDC electropherogram, are there any peaks significantly greater than 100 RFU? If not, proceed to next step. If there are any significant peaks in the NDC electropherogram, check whether the same peaks are also present in any other electropherograms. If any are, this suggests that one of the STR assay reagents may be contaminated with some human DNA. This is very unlikely to occur when using a reputable service facility using commercial STR assay kits, but may be more likely when using a facility that does not use a validated commercial kit. - 2. In the PDC genotype and your sample DNA electropherograms, check whether the peaks have been identified by allele identity and by allele size (sz) in base pair lengths. The intensities of the peaks are usually not important for cell line authentication, whereas the amplicon lengths can help in determining whether a very strong peak in one channel (pane) has spilled over into a different signal channel and thus generated a spurious peak. - 3. Examine the peaks in your sample(s). Are they properly identified by the above parameters (size and allele call)? If yes, proceed. If not, discuss your concerns with the service facility. If the peak sizes in base pairs are not indicated, use a ruler either on the computer screen or on the PDF printouts of the electropherograms to see whether any peaks spill over between spectral channels (panes). If there are such peaks with labeled sizes within 1 bp, these could be spillover peaks that should only be considered in the channel with the strongest signal. Note that the vertical scales in the different panes may be different when comparing peak intensities. The service facility should have normalized all the peaks relative to the strongest peak in each pane for optimal display of the data. - 4. Are there any spurious peaks that have been labeled as “OL” (off‐ladder)? If not, proceed. If there are such peaks, check for a very strong signal peak in a different channel that has the same amplicon size/length. If not, discuss your concerns with the service facility. If the peak base pair sizes are not indicated, use a ruler either on the computer screen or on the PDF printouts of the electropherograms to see whether any strong (high RFU >>20,000 RFU) peak has spilled over into one of the other panes. Data Analytical Steps Using CLASTR of Cellosaurus The purposes of the following steps are to determine whether the STR genotype corresponds to (1) the expected cell line and (2) the cell line you plan to use, as well as (3) whether the sample you are using is actually that of another cell line and not the sample you desire to use for your research (i.e., whether it has been cross‐contaminated). If the STR genotype is correct, you should proceed to create a large master bank of aliquots for use in the planned experiments. This is important because you will need to use fresh aliquots of your cell line after a maximum of every 10 passages for MSI‐unstable cell lines or 20 passages for MSI‐stable cell lines, and you will want all the samples to be from the same batch of cells. It is necessary to have a sufficient number of aliquots for your work and for any additional unplanned, but necessary, experimental repeats or subsequent tests. - 1. Examine the electropherograms for your samples, and use Table 4 of the accompanying protocol article (Korch et al., submitted) to evaluate the data quality and properties, beginning with the following questions and algorithms to test initially after loading the STR data in either CLASTR or in Appendix 2 of Korch et al. (submitted), as outlined below. Do all the STR loci have either one or two peaks? Use the TNB search algorithm to see whether there are any matches to your sample genotype. Do some loci have more than two peaks (alleles)? How many loci? If only one or two loci have more than two peaks, proceed with the TNB and MvR algorithms and evaluate the results. If several loci have two or more alleles, analyze with the MvR and TNB algorithms and examine the results. The MvR results may indicate whether the profile is composed of that from two or more cell lines. Do most or all of the loci have only a single peak? Use primarily the TNB and MvQ profile for analysis. - 2. STR data entry: If you have multiple genotypes to analyze, enter your new STR genotyping data into the Template for STR Profile CLASTR Search (.xlsx) from Appendix 1 of the accompanying protocol article (Korch et al., submitted). Alternatively, if you have only one or a few samples to analyze, you can enter the data individually in the online data entry form as described in Korch et al. (submitted) and outlined below. - 3. Go to CLASTR in Cellosaurus (https://www.cellosaurus.org/str‐search/) and either enter the data into the Similarity Search Tool or use the Load File option in the lower right side of the data entry form (Appendix 1 of Korch et al., submitted) to upload your data for comparison. This latter option is especially useful if you have multiple STR genotypes to enter. - 4. Submit the STR data for analysis. The analysis takes only a few seconds, even for multiple samples. The default analysis uses the TNB algorithm with filters set at a minimum score of 60% and a minimum number of 8 STR loci, and a maximum of 200 results. If needed these can be changed. The most likely change will be only which algorithm to use. - 5. The single data entry will produce a spreadsheet of results. In the upper left corner at the top of the table of results, there is an option to “Export table” with the results. The multiple genotype option using the “Load File” option will produce one file that has a separate worksheet for each submitted genotype in the Template for STR Profile CLASTR Search (.xlsx) from Appendix 1 of the accompanying protocol article (Korch et al., submitted) that was used to load the data. On completion of the analysis using the Load File option, you will be asked for a name and format (.csv or .xlsx) for the data. A handy naming format is a version which includes the following: date, (yyyymmdd), project name, cell line(s), and algorithm used. Use either option to download the results for review. - 6. Refer to Table 4 of Korch et al. (submitted) for analysis and troubleshooting of the STR results. Laboratory Database Setup This step is important for all future work with cell lines and tissue samples being used in the laboratory. It will allow you to follow cell line evolution in the laboratory and catch when any cell line has been cross‐contaminated by another cell line being used in the laboratory. - 1. Enter your new STR genotyping data into Worksheet 3 from Appendix 2 (ICLAC Human STR Genotype Macro Comparison Tool, .xlsm) of the accompanying protocol article (Korch et al., submitted). The STR data entered into this worksheet will serve as the laboratory reference genotypes for Worksheets 5 and 6 from Appendix 2 of Korch et al. (submitted). New data can be added to the worksheet as new genotypes are determined or obtained from other sources. Worksheet 3 from the same appendix can serve as the laboratory's STR genotype database. - 2. Proceed to Worksheets 5 and 6 and enter the STR genotyping data of the sample that you wish to query to determine whether there are any matches to any of the lab's other cell lines (those entered in Worksheet 3). - 3. Once the new sample has been queried for its identity against other cell lines in Worksheet 3, this genotype can be designated as a reference genotype and added to Worksheet 3 for future use. - 4. Repeat this process for each new sample to continue expanding the cell line genotype database. - 5. The resulting data in Worksheets 5 and 6 are shown in the order in which they are listed in Worksheet 3. Appendix 3 of Korch et al. (submitted) describes, with illustrations, the steps for sorting the data in Worksheets 5 and 6 of Appendix 2. The results can thus be ranked by their match scores using any of the three algorithms. IMPORTANT CELL LINE AUTHENTICATION CONSIDERATIONS Because the authenticity of cell lines and tissue samples is critical for valid and reproducible biomedical research, it is crucial to track sample identity during a project. This must be minimally done for each cell line at the five following stages of each project: - 1. Before starting to use a cell line in a project; - 2. Checking new cell samples that are growing in quarantine during one of the initial passages to determine whether they have the correct STR genotyped and are free of mycoplasma; - 3. After making the large master batch with multiple aliquots, of which one sample is used to confirm that it has the correct STR genotype; - 4. During the project: - a. When any unusual behavior or phenotype, even a desired one, is observed; - b. When any operator error is suspected, such as incorrect labeling or switching of samples; and - c. After any transfection and drug selection, because this process can inadvertently select and enrich for other unrelated cell lines being used in the laboratory. - a. - 5. At the end of a project, before submitting a grant application or a manuscript for publication. Remember that journals and reviewers may want to see the actual electropherograms for all the cell lines and tissue samples used for a manuscript. THE SCYLLA‐CHARYBDIS DILEMMA: PROTECTING PATIENT GENETIC INFORMATION CONFIDENTIALITY VERSUS THE NEED TO AUTHENTICATE CELL LINES AND IN VITRO CULTURED TISSUE SAMPLES BY STR GENOTYPING The original 8 or 13 forensic CODIS STR loci were selected so as not to reveal any physical or medical information about the patient/donor from which the sample was obtained. Most of these loci lie between genes in so‐called “gene deserts” or in intronic regions (Katsanis & Wagner, 2013). However, recent reviews note that some STR loci may be linked to genetic variants associated with phenotypes (Bañuelos et al., 2022; Wyner et al., 2020). Biomedical researchers have established cell lines from resected tumors and normal cells and tissues. The STR genotypes of DNA extracted from the original tissue or the patient's blood can then be compared to the DNA from the cell lines to establish the authenticity of the cultured cells. This procedure makes it possible to associate certain diseases due to germline and somatic mutations with specific patients, which raises the importance of protecting the sensitive protected health information (PHI) of patients to prevent violating regulations protecting the privacy of such information. The above analyses show that in many cases the 8, 9, or even 13 STR loci recommended in the ANSI‐ATCC ASN‐0002 standards of 2011 and 2022 are insufficient to specifically identify some cell lines and that optimal specificity requires 30 STR alleles. Some current STR kits are designed to test for alleles at 25 or more STR loci, including the original 8 or 13 CODIS loci. The use of such kits raises concerns, however, about identifying individuals and exposing their PHI. As discussed hitherto, cell line STR genotypes vary due to genetic instability, and consequently, the STR genotypes may not exactly match a unique person for forensic purposes, but STR genetic data should still be treated as sensitive PHI to ensure patient privacy. Key Regulatory Frameworks To address these issues, different countries and organizations have regulations governing the handling of STR data. In Japan, the Act on the Protection of Personal Information (APPI) protects genetic data and stipulates that only the alleles at 8 STR loci may be published. The M.D. Anderson Cancer Center in Texas, USA, has a similar, self‐imposed cell line policy about STR data. In Australia, there are rules governing the ethical use of human biospecimens, including cell lines. The European Union's General Data Protection Regulation (GDPR) is designed to protect genetic data, and the European Court of Human Rights (ECtHR) has ruled that STR data are protected under Article 8 of the European Convention on Human Rights (ECHR) for the right to respect patient privacy. The rules have allowed each country in the EU to specify the limits of what and how much genetic data could be published. In the United States, patient confidentiality and medical insurability are regulated by the Patient Protection and Affordable Care Act (AFA), the Health Insurance Portability and Accountability Act of 1996 (HIPAA), the Genetic Information Nondiscrimination Act (GINA), and the Americans with Disabilities Act (ADA). Genetic data, such as STR genotypes, are to be treated as sensitive PHI and therefore their publication is regulated by these different federal regulations. HIPAA restricts disclosure by “covered entities” without patient authorization, while GINA prohibits insurers and employers from using such data for discrimination. In academic laboratories, experiments with human samples are subject to regulation and oversight by internal review boards (IRBs). However, a 2013 decision by the US Supreme Court (Maryland vs. King) ruled that taking a DNA sample from an arrestee is permissible, especially for “serious” crimes (which were not defined), although the DNA results could not be used for purposes other than legal ones without explicit consent from all parties. This “search” permits the violation of the Fourth Amendment of the US Constitution if there is deemed to be “probable cause.” In 2026, the US Immigration and Customs Enforcement (ICE) division of the U.S. Department of Homeland Security began using Title 28, Chapter I, Part 28, Subpart B, § 28.12 of the US Code of Federal Regulations from 2020 to take DNA samples from anybody they stopped and took into custody, as reported by National Public Radio on March 19, 2026 (https://www.npr.org/transcripts/nx‐s1‐5739257). This means that the resulting DNA samples can be analyzed not only for identity purposes but also for other private information, although this would violate HIPAA and GINA regulations. In 2017, H.R. 1313 (the Preserving Employee Wellness Programs Act) was proposed in the U.S. House of Representatives with numerous exemptions to the GINA and ADA acts that would have allowed the violation of patient genetic confidentiality. Fortunately, this proposal was not enacted, but it raises the specter that economic and political forces may lead to attempts to reveal patient genetic information. To avoid exposing private information and violating the different privacy regulations, researchers must de‐identify (anonymize) STR data. Under HIPAA rules, STR data can be published without authorization only if it is stringently de‐identified, meaning that 18 specific identifiers (including names, dates, zip codes, addresses, social security numbers, medical and health plan numbers, biometric identifiers, face photographs, and vehicle identifiers) are removed and there is no reasonable basis to believe that the remaining data can identify an individual. For detailed listing and definition of these identifiers, see https://www.dhcs.ca.gov/dataandstats/data/Pages/ListofHIPAAIdentifiers.aspx. Rules for Publication Authorization requirement: Researchers or institutions publishing STR data linked to individuals must obtain signed, HIPAA‐compliant authorizations. Public data repositories: When uploading to public databases, data must be sufficiently anonymized, as de‐identified genetic data are still potentially re‐identifiable.

on employers and insurers: Employers are generally forbidden from disclosing genetic information except under very limited circumstances (e.g., court order, authorized workplace wellness programs). Exceptions: Authorization is not required for data to be disclosed to a public health agency for specific public health activities, or in response to a court order. Scope and Limitations Covered entities versus others: HIPAA and GINA apply primarily to covered entities (hospitals, insurers, doctors) and their business associates. Non‐covered entities: Companies such as direct‐to‐consumer genetic testing firms (e.g., 23andMe, Ancestry) or genealogy sites are not covered by HIPAA but only by their own in‐house privacy policies, potentially leaving data publication in these sectors with fewer protections. It is possible that a tool using STR data analogous to GEDmatch, which uses SNP data, might be able to identify a familial link between a patient and a cell line, as was done using SNP data to identify California's Golden State Killer (https://sequencing.com/blog/post/golden‐state‐killer‐dna‐kit‐privacy). But SNP testing, the procedure used for ancestry identification, usually tests for ≈600,000 SNPs and is much more useful than STR data for identifying members of a family tree. Re‐identification risks: Even with de‐identification, the high specificity of STR profiles carries risks of re‐identification, which remains a challenge under current regulatory frameworks (Lowrance & Collins, 2007) and as illustrated by the examples presented above. It has been shown that if extensive SNP data are available, they can be linked to STR data (Edge et al., 2017). Also, STR genotypes, along with disease‐causing mutations and physical appearance phenotypes, can be deduced from whole‐genome sequencing data, thus breaching the anonymity of anonymized personal data (Erlich & Narayanan, 2014; Erlich et al., 2014). It is not likely that the Fluidigm SNPtrace assay with 96 SNPs described earlier (Liang‐Chu et al., 2015) could provide sufficient information to link SNP data to STR data and other patient data. Another potential scenario by which PHI could be exposed is if, during a job interview, a candidate is asked for a blood sample for drug testing. Genomic DNA could be extracted from such a sample and used for both SNP and STR analyses, thus revealing genetic information about the applicant that was not the premise for originally taking the blood sample. The STR and SNP data could be used to screen for the existence of related cancerous cell lines, which could be disconcerting to the employer due to the possibility of the employee being prone to having cancer and thus raising the company's health insurance liabilities. Publication Recommendations We are not aware of any journal that limits the number of loci for which STR data can be reported. The International Journal of Cancer (IJC) requests data for a minimum of eight core STR loci (Souren et al., 2022; Wang et al., 2017). The Nature journals require authentication of cell lines, but do not specify the number of loci to be published; therefore, the decision of how much human STR genotyping data to publish rests with individual laboratories and research organizations. Souren et al. (2022) also recommended that only tabulated STR data, but not the actual electropherograms, be published. The electropherograms should be provided only to the editor and reviewers to use in evaluating manuscripts so that the quality of the STR data can be assessed. One reason for this stipulation is that the IJC has received several manuscripts in which the authors had copy‐and‐pasted electropherograms from articles published by other authors. Therefore, we recommend all scientists and research entities consider the above points when establishing their own policies regulating the release of human STR results and comply with their local and national regulations that govern the publication of these data. FINAL RECOMMENDATIONS AND USEFUL RESOURCES First, as indicated by this summary of the various types of STR data presented by cell lines, it is important to realize that cell lines are not stable, homogeneous, unchanging biological models. In Extended Data Table 1 of one publication (Ben‐David et al., 2018; https://www.nature.com/articles/s41586‐018‐0409‐3/tables/1), the authors summarized different causes of instability and variations in cell lines. They had observed that different samples of cell lines showed extensive genetic heterogeneity (both genomic and transcriptomic) due in part to passaging, genetic manipulations (even so‐called “neutral” changes), pre‐existing heterogeneity within cultures, and genetic instability due to MSI or TP53 mutational alterations. Cell line samples also present variability through misidentification due to various laboratory errors. Fortunately, version 14 (2026) of the ICLAC Registry of Misidentified Cell Lines (https://iclac.org/databases/cross‐contaminations/) lists 560 misidentified cell lines for which no known authentic samples have been identified. Vigorously growing cell lines such as HeLa, M14, MCF‐7, and others can easily take over cultures of other slower growing cells. HeLa cell lines can “fly”: that is, they can survive being dispersed in aerosols (Coriell, 1962) and thus infect other cell line cultures. For example, 90% of publications using the false cell line KB (CVCL_0372) describe it as being an oral epidermoid carcinoma cell line and not by its true identity as HeLa, a cervical adenocarcinoma (Vaughan et al., 2017). In the 1970s, Walter Nelson‐Rees identified about 280 imposter cell lines (Buehring, 2010). In 2017, 252 of ICLAC's list of false cell lines, of which there are no known authentic samples, were found to have been used in 32,755 published articles, which were subsequently cited in over half a million other articles (Horbach & Halffman, 2017). The HeLa‐derived cell line HEp‐2 was used under its false identity in 8497 articles (Gorphe, 2019; Korch & Capes‐Davis, 2021; Neimark, 2015). All cultures of cell lines can be easily mixed up and mislabeled. Examination of numerous studies of cell line identities, it has been estimated that about 2 of 9 (22%) cell line cultures are likely not to be correctly identified (Korch & Varella‐Garcia, 2018). To best address this huge problem prospectively, it is essential for research laboratories to implement a cell line and tissue authentication policy, as well as training on the use of cell lines. Consult at least the following six important resources that have compiled numerous recommendations for such training: Freshney's Culture of Animal Cells (Capes‐Davis & Freshney, 2021); ANSI‐ATCC ASN‐0002‐2022 (Korch et al., 2022); The assay guidance manual in the National Library of Medicine entitled Authentication of Human and Mouse Cell Lines by STR DNA Genotype Analysis (Almeida & Korch, 2023); The Cellosaurus website (https://www.cellosaurus.org); The International Cell Line Authentication Committee (ICLAC) website (https://iclac.org), especially the Resources section, with several guides for different aspects of cell line authentication (e.g., Institutional Best Laboratory Practices for Authentication of Cell Lines and Tissue Samples, Cell Line Checklist for Manuscripts and Grant Applications); Supporting Information Appendix 1 to this article, providing a list of “Twelve Steps for Safeguarding Biomedical Research Using Cell Lines and Tissue Samples” and a chart for tissue culture labs as a reminder of where culture handling pitfalls arise. The additional resources below are also useful: Rebecca Skloot, The immortal life of Henrietta Lacks (Skloot, 2010); Richard Harris, Rigor mortis (Harris, 2017); Michael Gold, A conspiracy of cells: One woman's immortal legacy and medical scandal it caused (Gold, 1986); Kevin Adam Curtis, The way of all flesh—The story about Henrietta Lacks and HeLa cells (Curtis, 1997); Biocompare, “Cell line misidentification and the problem of irreproducibility of biomedical research” (Biocompare, 2018; Oliver, 2018). In summary, this article and the accompanying cell line authentication protocol article (Korch et al., submitted) are targeted both at researchers and at journal editors and reviewers. They are designed to provide an understanding of the importance of cell line and tissue authentication, of how cell lines evolve with divergent handling, and of how different sublines can coexist in cultures of a cell lines that may differ both genetically and phenotypically. Although cell lines may not be perfect biological models, they can be used successfully for research if their limitations are acknowledged and understood and if they are checked for authenticity throughout a project and before results are submitted to granting agencies and journals. This is critical for preserving investigator and institutional reputations, avoiding the wasting of research funds, and stopping the pollution of the biomedical research literature with invalid and irreproducible results. Author Contributions Christopher T. Korch: Conceptualization; investigation; writing ‐ original draft; methodology; visualization; writing ‐ review and editing; validation; formal analysis; project administration; data curation; supervision; resources. Conflict of Interest The author is a member of the International Cell Line Authentication Committee (ICLAC), a group of biomedical researchers whose goals are (1) to identify and raise awareness among users of cell lines and tissue samples of the existence of mis‐identified/false/imposter cell lines, (2) to demonstrate how the cell line misidentification problem pollutes the biomedical literature with irreproducible and invalid research findings, (3) to identify spurious cell lines and publications based on such cell lines, and (4) to provide know‐how for tackling this problem. This study was self‐funded, and the author has no conflicts of interest, other than membership in ICLAC, which is a volunteer organization with no commercial funding. Supporting information Acknowledgments I wish to thank the following scientists for DNA samples and the numerous discussions about cell line authentication that have given me a deeper understanding of the need for cell line authentication, in particular: Adrian van Bokhoven, Gary Miller, Roland Nardone, Amanda Capes‐Davis, Richard Neve, Raymond Nims, Douglas Storts, Stanley Gartler, Rebecca Schweppe, Bryan Haugen, and Gregory Sykes. I am especially grateful to Rebecca Schweppe, with whom I have worked since 2007 on many aspects of keeping cell lines authentic and for sharing her lab space, and Amanda Capes‐Davis, for the many discussions since our first online meeting in 2012. I dedicate this work to the memory of Amos Bairoch, the developer of Cellosaurus, who will be greatly missed by the Cell Line Authentication Committee; Gary Miller, who started me down this path of authenticating cell lines; Roland Nardone, who revived the international concern about cell line authenticity and introduced me to ICLAC; and S. Richard Snow, for teaching me yeast genetics and guiding me through my PhD degree work. I am very grateful to the Divisions of Endocrinology, Metabolism and Diabetes and of Medical Oncology at the University of Colorado Anschutz Medical Campus for both laboratory and library access, which made this study possible. Korch, C. T. (2026). A biomedical researcher's guide for analyzing short tandem repeat (STR) genotypes of human cell lines and in Vitro tissue samples using three standard authentication algorithms. Current Protocols, 6, e70387. doi: 10.1002/cpz1.70387 Published in the Bioinformatics section Data Availability Statement The data that support the findings of this study are available in Cellosaurus at https://www.cellosaurus.org/. Literature Cited - Albitar, L. , Pickett, G. , Morgan, M. , Davies, S. , & Leslie, K. K. (2007). Models representing type I and type II human endometrial cancers: Ishikawa H and Hec50co cells. Gynecologic Oncology, 106(1), 52–64. 10.1016/j.ygyno.2007.02.033 [DOI] [PubMed] [Google Scholar] - Almeida, J. L. , Dakic, A. , Kindig, K. , Kone, M. , Letham, D. L. D. , Langdon, S. , Peat, R. , Holding‐Pillai, J. , Hall, E. M. , Ladd, M. , Shaffer, M. D. , Berg, H. , Li, J. , Wigger, G. , Lund, S. , Steffen, C. R. , Fransway, B. B. , Geraghty, B. , Natoli, M. , … Reid, Y. (2019). Interlaboratory study to validate a STR profiling method for intraspecies identification of mouse cell lines. PLoS ONE, 4(6), e0218412. 10.1371/journal.pone.0218412 [DOI] [PMC free article] [PubMed] [Google Scholar] - Almeida, J. L. , Hill, C. R. , & Cole, K. D. (2014). Mouse cell line authentication. Cytotechnology, 66(1), 133–147. 10.1007/s10616-013-9545-7 [DOI] [PMC free article] [PubMed] [Google Scholar] - Almeida, J. L. , & Korch, C. T. (2023). Authentication of human and mouse cell lines by short tandem repeat (STR) DNA genotype analysis. In Markossian, S. , Grossman, A. , Baskir, H. , Arkin, M. , Auld, D. , Austin, C. , Baell, J. , Brimacombe, K. , Chung, T. D. Y. , Coussens, N. P. , Dahlin, J. L. , Devanarayan, V. , Foley, T. L. , Glicksman, M. , Gorshkov, K. , Grotegut, S. , Hall, M. D. , Hoare, S. , Inglese, J. , Iversen, P. W. , Lal‐Nag, M. , Li, Z. , Manro, J. R. , McGee, J. , Norvil, A. , Pearson, M. , Riss, T. , Saradjian, P. , Sittampalam, G. S. , Tarselli, M. A. , Trask Jr., O. J. , Weidner, J. R. , Wildey, M. J. , Wilson, K. , Xia, M. , & Xu, X. (Eds.), The assay guidance manual. National Center for Advancing Translational Sciences. https://www.ncbi.nlm.nih.gov/books/NBK144066/pdf/Bookshelf_NBK144066.pdf [PubMed] [Google Scholar] - Bady, P. , Diserens, A. C. , Castella, V. , Kalt, S. , Heinimann, K. , Hamou, M. F. , Delorenzi, M. , & Hegi, M. E. (2012). DNA fingerprinting of glioma cell lines and considerations on similarity measurements. Neuro‐Oncology, 4(6), 701–711. 10.1093/neuonc/nos072 [DOI] [PMC free article] [PubMed] [Google Scholar] - Bañuelos, M. M. , Zavaleta, Y. J. A. , Roldan, A. , Reyes, R. J. , Guardado, M. , Chavez Rojas, B. , Nyein, T. , Rodriguez Vega, A. , Santos, M. , Huerta‐Sanchez, E. , & Rohlfs, R. V. (2022). Associations between forensic loci and expression levels of neighboring genes may compromise medical privacy. Proceedings of the National Academy of Sciences of the United States of America, 119(40), e2121024119. 10.1073/pnas.2121024119 [DOI] [PMC free article] [PubMed] [Google Scholar] - Ben‐David, U. , Siranosian, B. , Ha, G. , Tang, H. , Oren, Y. , Hinohara, K. , Strathdee, C. A. , Dempster, J. , Lyons, N. J. , Burns, R. , Nag, A. , Kugener, G. , Cimini, B. , Tsvetkov, P. , Maruvka, Y. E. , O'Rourke, R. , Garrity, A. , Tubelli, A. A. , Bandopadhayay, P. , … Golub, T. R. (2018). Genetic and transcriptional evolution alters cancer cell line drug response. Nature, 560(7718), 325–330. 10.1038/s41586-018-0409-3 [DOI] [PMC free article] [PubMed] [Google Scholar] - Berchuck, A. , & Boyd, J. (1995). Molecular basis of endometrial cancer. Cancer, 76(10 Suppl), 2034–2040. 10.1002/1097-0142(19951115)76:10+[2034::aid-cncr2820761321^3.0.co;2-u [DOI] [PubMed] [Google Scholar] - Biocompare . (2018). Reproducibility issues in life science research: The importance of cell line authentication [video]. Biocompare. https://www.biocompare.com/Reproducibility/Cell‐Line‐Authentication/ - Buehring, G. C. (2010). In Memoriam. Walter Anthony Nelson‐Rees, Research Geneticist, UC Berkeley, 1929–2009. Retrieved from http://senate.universityofcalifornia.edu/_files/inmemoriam/html/walteranthonynelson‐rees.html - Butler, J. M. , Vallone, P. M. , Gettings, K. B. , Borsuk, L. A. , Ruitberg, C. M. , & Reeder, D. J. (2025). NIST Short Tandem Repeat DNA Internet Database, Version 2.0. Retrieved from https://strbase.nist.gov - Capes‐Davis, A. , & Freshney, R. I. (2021). Freshney's culture of animal cells: A manual of basic technique and specialized applications (8th ed.). Wiley‐Blackwell. [Google Scholar] - Capes‐Davis, A. , Reid, Y. A. , Kline, M. C. , Storts, D. R. , Strauss, E. , Dirks, W. G. , Drexler, H. G. , MacLeod, R. A. , Sykes, G. , Kohara, A. , Nakamura, Y. , Elmore, E. , Nims, R. W. , Alston‐Roberts, C. , Barallon, R. , Los, G. V. , Nardone, R. M. , Price, P. J. , Steuer, A. , … Kerrigan, L. (2013). Match criteria for human cell line authentication: Where do we draw the line? International Journal of Cancer, 132(11), 2510–2519. 10.1002/ijc.27931 [DOI] [PubMed] [Google Scholar] - Castro, F. , Dirks, W. G. , Fahnrich, S. , Hotz‐Wagenblatt, A. , Pawlita, M. , & Schmitt, M. (2013). High‐throughput SNP‐based authentication of human cell lines. International Journal of Cancer, 132(2), 308–314. 10.1002/ijc.27675 [DOI] [PMC free article] [PubMed] [Google Scholar] - Chan, A. C. , Martyn, G. D. , & Carter, P. J. (2025). Fifty years of monoclonals: The past, present and future of antibody therapeutics. Nature Reviews Immunology, 25(10), 745–765. 10.1038/s41577-025-01207-9 [DOI] [PubMed] [Google Scholar] - Chesnokova, L. S. , Nishimura, S. L. , & Hutt‐Fletcher, L. M. (2009). Fusion of epithelial cells by Epstein‐Barr virus proteins is triggered by binding of viral glycoproteins gHgL to integrins alphavbeta6 or alphavbeta8. Proceedings of the National Academy of Sciences of the United States of America, 106(48), 20464–20469. 10.1073/pnas.0907508106 [DOI] [PMC free article] [PubMed] [Google Scholar] - Cooper, J. K. , Sykes, G. , King, S. , Cottrill, K. , Ivanova, N. V. , Hanner, R. , & Ikonomi, P. (2007). Species identification in cell culture: A two‐pronged molecular approach. In Vitro Cellular & Developmental Biology–Animal, 43(10), 344–351. 10.1007/s11626-007-9060-2 [DOI] [PubMed] [Google Scholar] - Coriell, L. L. (1962). Detection and elimination of contaminating organisms. National Cancer Institute Monograph, 7, 33–53. [PubMed] [Google Scholar] - Curtis, A. (1997). The way of all flesh—The story about Henrietta Lacks and HeLa cells. In Curtis (producer) A., Modern Time. British Broadcasting Corporation. https://www.bbc.co.uk/programmes/p08mqggg [Google Scholar] - Devor, E. J. , Gonzalez‐Bosquet, J. , Thiel, K. W. , & Leslie, K. K. (2020). Genomic characterization of five commonly used endometrial cancer cell lines. International Journal of Oncology, 57(6), 1348–1357. 10.3892/ijo.2020.5139 [DOI] [PMC free article] [PubMed] [Google Scholar] - Dice, L. R. (1945). Measures of the amount of ecologic association between species. Ecology, 26(3), 297–302. 10.2307/1932409 [DOI] [Google Scholar] - Edge, M. D. , Algee‐Hewitt, B. F. B. , Pemberton, T. J. , Li, J. Z. , & Rosenberg, N. A. (2017). Linkage disequilibrium matches forensic genetic records to disjoint genomic marker sets. Proceedings of the National Academy of Sciences of the United States of America, 114(22), 5671–5676. 10.1073/pnas.1619944114 [DOI] [PMC free article] [PubMed] [Google Scholar] - Eltonsy, N. , Gabisi, V. , Li, X. , Russe, K. B. , Mills, G. B. , & Stemke‐Hale, K. (2012). Detection algorithm for the validation of human cell lines. International Journal of Cancer, 131(6), E1024–E1030. 10.1002/ijc.27533 [DOI] [PMC free article] [PubMed] [Google Scholar] - Ephrussi, B. , & Weiss, M. C. (1965). Interspecific hybridization of somatic cells. Proceedings of the National Academy of Sciences of the United States of America, 53(5), 1040–1042. 10.1073/pnas.53.5.1040 [DOI] [PMC free article] [PubMed] [Google Scholar] - Erlich, Y. , & Narayanan, A. (2014). Routes for breaching and protecting genetic privacy. Nature Reviews Genetics, 15(6), 409–421. 10.1038/nrg3723 [DOI] [PMC free article] [PubMed] [Google Scholar] - Erlich, Y. , Williams, J. B. , Glazer, D. , Yocum, K. , Farahany, N. , Olson, M. , Narayanan, A. , Stein, L. D. , Witkowski, J. A. , & Kain, R. C. (2014). Redefining genomic privacy: Trust and empowerment. PLoS Biology, 12(11), e1001983. 10.1371/journal.pbio.1001983 [DOI] [PMC free article] [PubMed] [Google Scholar] - Gartler, S. M. (1967). Genetic markers as tracers in cell culture. National Cancer Institute Monograph, 26, 167–195. [PubMed] [Google Scholar] - Gartler, S. M. (1968). Apparent HeLa cell contamination of human heteroploid cell lines. Nature, 217(5130), 750–751. 10.1038/217750a0 [DOI] [PubMed] [Google Scholar] - Gettings, K. B. , Borsuk, L. A. , Steffen, C. R. , Kiesler, K. M. , & Vallone, P. M. (2018). Sequence‐based U.S. population data for 27 autosomal STR loci. Forensic Science International: Genetics, 7, 106–115. 10.1016/j.fsigen.2018.07.013 [DOI] [PMC free article] [PubMed] [Google Scholar] - Gold, M. (1986). A conspiracy of cells—One woman's immortal legacy and the medical scandal it caused. State University of New York Press. [Google Scholar] - Gorphe, P. (2019). A comprehensive review of Hep‐2 cell line in translational research for laryngeal cancer. American Journal of Cancer Research, 9(4), 644–649. [PMC free article] [PubMed] [Google Scholar] - Han, S. J. , Hawkins, S. M. , Begum, K. , Jung, S. Y. , Kovanci, E. , Qin, J. , Lydon, J. P. , DeMayo, F. J. , & O'Malley, B. W. (2012). A new isoform of steroid receptor coactivator‐1 is crucial for pathogenic progression of endometriosis. Nature Medicine, 18(7), 1102–1111. 10.1038/nm.2826 [DOI] [PMC free article] [PubMed] [Google Scholar] - Harris, H. , & Watkins, J. F. (1965). Hybrid cells derived from mouse and man: Artificial heterokaryons of mammalian cells from different species. Nature, 205, 640–646. 10.1038/205640a0 [DOI] [PubMed] [Google Scholar] - Harris, R. (2017). Rigor mortis—How sloppy science creates worthless cures, crushes hope, and wastes billions. Basic Books. [Google Scholar] - Hombach‐Klonisch, S. , Kehlen, A. , Fowler, P. A. , Huppertz, B. , Jugert, J. F. , Bischoff, G. , Schlüter, E. , Buchmann, J. , & Klonisch, T. (2005). Regulation of functional steroid receptors and ligand‐induced responses in telomerase‐immortalized human endometrial epithelial cells. Journal of Molecular Endocrinology, 34(2), 517–534. 10.1677/jme.1.01550 [DOI] [PubMed] [Google Scholar] - Horbach, S. P. K. M. , & Halffman, W. (2017). The ghosts of HeLa: How cell line misidentification contaminates the scientific literature. PLoS ONE, 12(10), e0186281. 10.1371/journal.pone.0186281 [DOI] [PMC free article] [PubMed] [Google Scholar] - Huang, Y. , Liu, Y. , Zheng, C. , & Shen, C. (2017). Investigation of cross‐contamination and misidentification of 278 widely used tumor cell lines. PLoS ONE, 12(1), e0170384. 10.1371/journal.pone.0170384 [DOI] [PMC free article] [PubMed] [Google Scholar] - Jacobsen, B. M. , Harrell, J. C. , Jedlicka, P. , Borges, V. F. , Varella‐Garcia, M. , & Horwitz, K. B. (2006). Spontaneous fusion with, and transformation of mouse stroma by, malignant human breast cancer epithelium. Cancer Research, 66(16), 8274–8279. 10.1158/0008-5472.can-06-1456 [DOI] [PubMed] [Google Scholar] - Katsanis, S. H. , & Wagner, J. K. (2013). Characterization of the standard and recommended CODIS markers. Journal of Forensic Sciences, 58(Suppl 1), S169–S172. 10.1111/j.1556-4029.2012.02253.x [DOI] [PMC free article] [PubMed] [Google Scholar] - Kleensang, A. , Vantangoli, M. M. , Odwin‐DaCosta, S. , Andersen, M. E. , Boekelheide, K. , Bouhifd, M. , Fornace, A. J. , Jr, L. , H, H. , Livi, C. B. , Madnick, S. , Maertens, A. , Rosenberg, M. , Yager, J. D. , Zhao, L. , & Hartung, T. (2016). Genetic variability in a frozen batch of MCF‐7 cells invisible in routine authentication affecting cell function. Scientific Reports, 6, 28994. 10.1038/srep28994 [DOI] [PMC free article] [PubMed] [Google Scholar] - Köhler, G. , & Milstein, C. (1975). Continuous cultures of fused cells secreting antibody of predefined specificity. Nature, 256(5517), 495–497. 10.1038/256495a0 [DOI] [PubMed] [Google Scholar] - Korch, C. , & Capes‐Davis, A. (2021). The extensive and expensive impacts of HEp‐2 [HeLa], intestine 407 [HeLa], and other false cell lines in journal publications. SLAS Discovery, 26(10), 1268–1279. 10.1177/24725552211051963 [DOI] [PubMed] [Google Scholar] - Korch, C. , Capes‐Davis, A. , & Shen, C. (submitted) Analytical workflow protocol for determining the authenticity of human cell lines and tissue samples by STR genotype analysis. Current Protocols. [Google Scholar] - Korch, C. , Hall, E. M. , Dirks, W. G. , Ewing, M. , Faries, M. , Varella‐Garcia, M. , Robinson, S. , Storts, D. , Turner, J. A. , Wang, Y. , Burnett, E. C. , Healy, L. , Kniss, D. , Neve, R. M. , Nims, R. W. , Reid, Y. A. , Robinson, W. A. , & Capes‐Davis, A. (2018). Authentication of M14 melanoma cell line proves misidentification of MDA‐MB‐435 breast cancer cell line. International Journal of Cancer, 142(3), 561–572. 10.1002/ijc.31067 [DOI] [PMC free article] [PubMed] [Google Scholar] - Korch, C. , Hall, E. M. , Dirks, W. G. , Sykes, G. R. , Capes‐Davis, A. , Butler, J. M. , … Nardone, R. M. (2022). Human Cell Line Authentication. Standardization of Short Tandem Repeat (STR) Profiling, ASN‐0002, Revised 2022 (November 2022 ed.). American National Standards Institute (ANSI)–American Type Culture Collection (ATCC) Standards Development Organization. [Google Scholar] - Korch, C. , Spillman, M. A. , Jackson, T. A. , Jacobsen, B. M. , Murphy, S. K. , Lessey, B. A. , Jordan, V. C. , & Bradford, A. P. (2012). DNA profiling analysis of endometrial and ovarian cell lines reveals misidentification, redundancy and contamination. Gynecologic Oncology, 127(1), 241–248. 10.1016/j.ygyno.2012.06.017 [DOI] [PMC free article] [PubMed] [Google Scholar] - Korch, C. , & Varella‐Garcia, M. (2018). Tackling the human cell line and tissue misidentification problem is needed for reproducible biomedical research. Advances in Molecular Pathology, 1(1), 209–228.e36. 10.1016/j.yamp.2018.07.003 [DOI] [Google Scholar] - Lee, H. Y. , Park, M. J. , Yoo, J. E. , Chung, U. , Han, G. R. , & Shin, K. J. (2005). Selection of twenty‐four highly informative SNP markers for human identification and paternity analysis in Koreans. Forensic Science International, 148(2–3), 107–112. 10.1016/j.forsciint.2004.04.073 [DOI] [PubMed] [Google Scholar] - Li, Y. , Arao, Y. , Hall, J. M. , Burkett, S. , Liu, L. , Gerrish, K. , Cavailles, V. , & Korach, K. S. (2014). Research resource: STR DNA profile and gene expression comparisons of human BG‐1 cells and a BG‐1/MCF‐7 clonal variant. Molecular Endocrinology, 28(12), 2072–2081. 10.1210/me.2014-1229 [DOI] [PMC free article] [PubMed] [Google Scholar] - Liang‐Chu, M. M. , Yu, M. , Haverty, P. M. , Koeman, J. , Ziegle, J. , Lee, M. , Bourgon, R. , & Neve, R. M. (2015). Human biosample authentication using the high‐throughput, cost‐effective SNPtrace(TM) system. PLoS ONE, 10(2), e0116218. 10.1371/journal.pone.0116218 [DOI] [PMC free article] [PubMed] [Google Scholar] - Lorenzi, P. L. , Reinhold, W. C. , Varma, S. , Hutchinson, A. A. , Pommier, Y. , Chanock, S. J. , & Weinstein, J. N. (2009). DNA fingerprinting of the NCI‐60 cell line panel. Molecular Cancer Therapeutics, 8(4), 713–724. 10.1158/1535-7163.Mct-08-0921 [DOI] [PMC free article] [PubMed] [Google Scholar] - Lowrance, W. W. , & Collins, F. S. (2007). Ethics. Identifiability in genomic research. Science, 317(5838), 600–602. 10.1126/science.1147699 [DOI] [PubMed] [Google Scholar] - Masters, J. R. , Thomson, J. A. , Daly‐Burns, B. , Reid, Y. A. , Dirks, W. G. , Packer, P. , Toji, L. H. , Ohno, T. , Tanabe, H. , Arlett, C. F. , Kelland, L. R. , Harrison, M. , Virmani, A. , Ward, T. H. , Ayres, K. L. , & Debenham, P. G. (2001). Short tandem repeat profiling provides an international reference standard for human cell lines. Proceedings of the National Academy of Sciences of the United States of America, 98(14), 8012–8017. 10.1073/pnas.121616198 [DOI] [PMC free article] [PubMed] [Google Scholar] - Neimark, J. (2014). Trial and error—The dirty little secret of cancer research. Discover Magazine, (November), 44–51. [Google Scholar] - Neimark, J. (2015). Line of attack. Science, 347(6225), 938–940. 10.1126/science.347.6225.938 [DOI] [PubMed] [Google Scholar] - Nims, R. W. , Capes‐Davis, A. , Korch, C. , & Reid, Y. A. (2018). Authenticating hybrid cell lines. In Mehanna R. A. (Ed.), Cell Culture. IntechOpen Limited. Retrieved from https://www.intechopen.com/online‐first/authenticating‐hybrid‐cell‐lines. 10.5772/intechopen.80669 [DOI] [Google Scholar] - Nishida, M. (2002). The Ishikawa cells from birth to the present. Human Cell, 15(3), 104–117. [DOI] [PubMed] [Google Scholar] - Oliver, T. (2018). Combating misidentified and contaminated cell lines. Life Science Articles, Editorial Article, Retrieved from https://www.biocompare.com/Editorial‐Articles/348977‐Authenticate‐Combating‐Misidentified‐and‐Contaminated‐Cell‐Lines/ [Google Scholar] - Parson, W. , Kirchebner, R. , Muhlmann, R. , Renner, K. , Kofler, A. , Schmidt, S. , & Kofler, R. (2005). Cancer cell line identification by short tandem repeat profiling: Power and limitations. The FASEB Journal, 19(3), 1–18. 10.1096/fj.04-3062fje [DOI] [PubMed] [Google Scholar] - Romano, P. , Manniello, A. , Aresu, O. , Armento, M. , Cesaro, M. , & Parodi, B. (2009). Cell Line Data Base: Structure and recent improvements towards molecular authentication of human cell lines. Nucleic Acids Research, 37, (Database issue), D925–D932. 10.1093/nar/gkn730 [DOI] [PMC free article] [PubMed] [Google Scholar] - Romano, P. D. , Visconti, P. , & Parodi, B. (2016). Recent developments of the cell line integrated molecular authentication database. PeerJ Preprints, Database, 4, e2194v2191. 10.7287/peerj.preprints.2194v1 [DOI] [Google Scholar] - Ruitberg, C. M. , Reeder, D. J. , & Butler, J. M. (2001). STRBase: A short tandem repeat DNA database for the human identity testing community. Nucleic Acids Research, 29(1), 320–322. [DOI] [PMC free article] [PubMed] [Google Scholar] - Satyaswaroop, P. G. , & Tabibzadeh, S. S. (1991). Extracellular matrix and the patterns of differentiation of human endometrial carcinomas in vitro and in vivo. Cancer Research, 51(20), 5661–5666. [PubMed] [Google Scholar] - Saxon, P. J. , Srivatsan, E. S. , Leipzig, G. V. , Sameshima, J. H. , & Stanbridge, E. J. (1985). Selective transfer of individual human chromosomes to recipient cells. Molecular and Cellular Biology, 5(1), 140–146. [DOI] [PMC free article] [PubMed] [Google Scholar] - Saxton, R. E. , Irie, R. F. , Ferrone, S. , Pellegrino, M. A. , & Morton, D. L. (1978). Establishment of paired tumor cells and autologous virus‐transformed cell lines to define humoral immune responses in melanoma and sarcoma patients. International Journal of Cancer, 21(3), 299–306. 10.1002/ijc.2910210308 [DOI] [PubMed] [Google Scholar] - Schweppe, R. E. , Klopper, J. P. , Korch, C. , Pugazhenthi, U. , Benezra, M. , Knauf, J. A. , Fagin, J. A. , Marlow, L. A. , Copland, J. A. , Smallridge, R. C. , & Haugen, B. R. (2008). Deoxyribonucleic acid profiling analysis of 40 human thyroid cancer cell lines reveals cross‐contamination resulting in cell line redundancy and misidentification. The Journal of Clinical Endocrinology & Metabolism, 93(11), 4331–4341. 10.1210/jc.2008-1102 [DOI] [PMC free article] [PubMed] [Google Scholar] - Skloot, R. (2010). The immortal life of Henrietta Lacks. Crown Publishers. [Google Scholar] - Soleymani, A. , Hatami, N. , Jamkhane, A. E. , & Shabankhani, K. (2026). Prognostic and predictive roles of microsatellite instability and mismatch repair deficiency in early‐stage endometrial cancer: A structured narrative review. Obstetrics & Gynecology Science, 69(1), 27–38. 10.5468/ogs.25299 [DOI] [PMC free article] [PubMed] [Google Scholar] - Somaschini, A. , Amboldi, N. , Nuzzo, A. , Scacheri, E. , Ukmar, G. , Ballinari, D. , Malyszko, J. , Raddrizzani, L. , Landonio, A. , Gasparri, F. , Galvani, A. , Isacchi, A. , & Bosotti, R. (2013). Cell line identity finding by fingerprinting, an optimized resource for short tandem repeat profile authentication. Genetic Testing and Molecular Biomarkers, 17(3), 254–259. 10.1089/gtmb.2012.0359 [DOI] [PubMed] [Google Scholar] - Sørensen, T. (1948). A method of establishing groups of equal amplitudes in plant sociology based on similarity of species content and its application to analyses of the vegetation on Danish commons. Kongelige Danske Videnskabernes Selskab. Det Kongelige Danske Videnskabernes Selskab Biologiske Skrifter, 5(4), 1–34. [Google Scholar] - Souren, N. Y. , Fusenig, N. E. , Heck, S. , Dirks, W. G. , Capes‐Davis, A. , Bianchini, F. , & Plass, C. (2022). Cell line authentication: A necessity for reproducible biomedical research. The EMBO Journal, 41(14), e111307. 10.15252/embj.2022111307 [DOI] [PMC free article] [PubMed] [Google Scholar] - Steffen, C. R. , Coble, M. D. , Gettings, K. B. , & Vallone, P. M. (2017). Corrigendum to 'U.S. Population Data for 29 Autosomal STR Loci' [Forensic Sci. Int. Genet. 7 (2013) e82–e83]. Forensic Science International: Genetics, 31, e36–e40. 10.1016/j.fsigen.2017.08.011 [DOI] [PubMed] [Google Scholar] - Strong, M. J. , Baddoo, M. , Nanbo, A. , Xu, M. , Puetter, A. , & Lin, Z. (2014). Comprehensive high‐throughput RNA sequencing analysis reveals contamination of multiple nasopharyngeal carcinoma cell lines with HeLa cell genomes. Journal of Virology, 88(18), 10696–10704. 10.1128/jvi.01457-14 [DOI] [PMC free article] [PubMed] [Google Scholar] - Tabibzadeh, S. , Kaffka, K. L. , Kilian, P. L. , & Satyaswaroop, P. G. (1990). Human endometrial epithelial cell lines for studying steroid and cytokine actions. In Vitro Cellular & Developmental Biology, 26(12), 1173–1179. 10.1007/bf02623695 [DOI] [PubMed] [Google Scholar] - Tanabe, H. , Takada, Y. , Minegishi, D. , Kurematsu, M. , Masui, T. , & Mizusawa, H. (1999). Cell line individualization by STR multiplex system in the cell bank found cross‐contamination between ECV304 and EJ‐1/T24. Tissue Culture Research Communications, Retrieved from https://www.jstage.jst.go.jp/article/jtca1981/18/4/18_329/_article/‐char/en [Google Scholar] - van Bokhoven, A. , Varella‐Garcia, M. , Korch, C. , Hessels, D. , & Miller, G. J. (2001). Widely used prostate carcinoma cell lines share common origins. The Prostate, 47(1), 36–51. 10.1002/pros.1045 [DOI] [PubMed] [Google Scholar] - van Bokhoven, A. , Varella‐Garcia, M. , Korch, C. , Johannes, W. U. , Smith, E. E. , Miller, H. L. , Nordeen, S. K. , Miller, G. J. , & Lucia, M. S. (2003). Molecular characterization of human prostate carcinoma cell lines. The Prostate, 57(3), 205–225. 10.1002/pros.10290 [DOI] [PubMed] [Google Scholar] - Vaughan, L. , Glanzel, W. , Korch, C. , & Capes‐Davis, A. (2017). Widespread use of misidentified cell line KB (HeLa): Incorrect attribution and its impact revealed through mining the scientific literature. Cancer Research, 77(11), 2784–2788. 10.1158/0008-5472.can-16-2258 [DOI] [PubMed] [Google Scholar] - Vermeulen, S. J. , Chen, T. R. , Speleman, F. , Nollet, F. , van Roy, F. M. , & Mareel, M. M. (1998). Did the four human cancer cell lines DLD‐1, HCT‐15, HCT‐8, and HRT‐18 originate from one and the same patient? Cancer Genetics and Cytogenetics, 107(1), 76–79. 10.1016/s0165-4608(98)00081-8 [DOI] [PubMed] [Google Scholar] - Wang, M. , Wu, C. , Guo, Y. , Cao, X. , Zheng, W. , & Fan, G. K. (2017). The primary growth of laryngeal squamous cell carcinoma cells in vitro is effectively supported by paired cancer‐associated fibroblasts alone. Tumor Biology, 39(5), 101042831770551. 10.1177/1010428317705512 [DOI] [PubMed] [Google Scholar] - Weiskirchen, R. (2026). Genetic insights into the economic toll of cell line misidentification: A comprehensive review. Medical Sciences, 14(1), 25. 10.3390/medsci14010025 [DOI] [PMC free article] [PubMed] [Google Scholar] - Weiss, M. C. , & Green, H. (1967). Human‐mouse hybrid cell lines containing partial complements of human chromosomes and functioning human genes. Proceedings of the National Academy of Sciences of the United States of America, 58(3), 1104–1111. 10.1073/pnas.58.3.1104 [DOI] [PMC free article] [PubMed] [Google Scholar] - Willing, C. , Peich, M. , Danescu, A. , Kehlen, A. , Fowler, P. A. , & Hombach‐Klonisch, S. (2011). Estrogen‐independent actions of environmentally relevant AhR‐agonists in human endometrial epithelial cells. Molecular Human Reproduction, 17(2), 115–126. 10.1093/molehr/gaq081 [DOI] [PubMed] [Google Scholar] - Willing, C. , Peich, M. , Danescu, A. , Kehlen, A. , Fowler, P. A. , & Hombach‐Klonisch, S. (2013). Estrogen‐independent actions of environmentally relevant AhR‐agonists in human endometrial epithelial cells. MHR: Basic Science of Reproductive Medicine, 19(7), 473–473. 10.1093/molehr/gat035 [DOI] [PubMed] [Google Scholar] - Wyner, N. , Barash, M. , & McNevin, D. (2020). Forensic autosomal short tandem repeats and their potential association with phenotype. Frontiers in Genetics, 11, 884. 10.3389/fgene.2020.00884 [DOI] [PMC free article] [PubMed] [Google Scholar] - Ye, F. , Chen, C. , Qin, J. , Liu, J. , & Zheng, C. (2015). Genetic profiling reveals an alarming rate of cross‐contamination among human cell lines used in China. The FASEB Journal, 29(10), 4268–4272. 10.1096/fj.14-266718 [DOI] [PubMed] [Google Scholar] - Yu, M. , Selvaraj, S. K. , Liang‐Chu, M. M. , Aghajani, S. , Busse, M. , Yuan, J. , Lee, G. , Peale, F. , Klijn, C. , Bourgon, R. , Kaminker, J. S. , & Neve, R. M. (2015). A resource for cell line authentication, annotation and quality control. Nature, 520(7547), 307–311. 10.1038/nature14397 [DOI] [PubMed] [Google Scholar] - Zhao, M. , Sano, D. , Pickering, C. R. , Jasser, S. A. , Henderson, Y. C. , Clayman, G. L. , Sturgis, E. M. , Ow, T. J. , Lotan, R. , Carey, T. E. , Sacks, P. G. , Grandis, J. R. , Sidransky, D. , Heldin, N. E. , & Myers, J. N. (2011). Assembly and initial characterization of a panel of 85 genomically validated cell lines from diverse head and neck tumor sites. Clinical Cancer Research, 17(23), 7248–7264. 10.1158/1078-0432.Ccr-11-0690 [DOI] [PMC free article] [PubMed] [Google Scholar] Associated Data This section collects any data citations, data availability statements, or supplementary materials included in this article. Supplementary Materials Data Availability Statement The data that support the findings of this study are available in Cellosaurus at https://www.cellosaurus.org/.