Accurate prediction of colorectal cancer diagnosis using machine learning based on immunohistochemistry pathological images

Accurate prediction of colorectal cancer diagnosis using machine learning based on immunohistochemistry pathological images | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Article Accurate prediction of colorectal cancer diagnosis using machine learning based on immunohistochemistry pathological images Bobin Ning, Jimei Chi, Qingyu Meng, Baoqing Jia This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-4129792/v1 This work is licensed under a CC BY 4.0 License Status: Published Journal Publication published 02 Dec, 2024 Read the published version in Scientific Reports → Version 1 posted 12 You are reading this latest preprint version Abstract Colorectal cancer (CRC) ranks as the third most prevalent tumor and the second leading cause of mortality. Early and accurate diagnosis holds significant importance in enhancing patient treatment and prognosis. Machine learning technology and bioinformatics have provided novel approaches for cancer diagnosis. This study aims to develop a CRC diagnostic model based on immunohistochemical staining image features using machine learning methods. Initially, CRC disease-specific genes were identified through bioinformatics analysis and Random Forest algorithm utilizing RNA-seq data from both GEO and TCGA databases. Subsequently, verification of these genes was performed using proteomics data from CPTAC and HPA database, resulting in identification of target proteins (AKR1B10, CA2, DHRS9, and ZG16) for further investigation. SVM algorithm was then employed to analyze and integrate the characteristics of immunohistochemical images to construct a reliable CRC diagnostic model. During the training and validation process of this model, cross-validation along with external validation methods were implemented to ensure accuracy and reliability. The results demonstrate that the established diagnostic model exhibits excellent performance in distinguishing between CRC and normal controls (accuracy rate: 0.999), thereby presenting potential prospects for clinical application. These findings are expected to provide innovative perspectives as well as methodologies for personalized diagnosis of CRC while offering more precise references for promising treatment. Biological sciences/Cancer Biological sciences/Computational biology and bioinformatics Health sciences/Biomarkers Health sciences/Gastroenterology Health sciences/Oncology colorectal cancer diagnosis machine learning immunohistochemistry Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Introduction Colorectal cancer (CRC) is a prevalent malignant tumor, and its incidence is progressively increasing due to the changing unhealthy lifestyle habits, dietary patterns, and aging [ 1 ] . It primarily affects individuals over 50 years old [ 2 ] . Despite advancements in early screening and treatment methods that have improved patient survival rates, CRC remains one of the leading causes of death worldwide [ 3 ] . Many patients are diagnosed at an advanced stage, which significantly increases the risk of recurrence and metastasis [ 4 ] . Therefore, timely diagnosis plays a crucial role for CRC patients as it enables early detection of the disease's presence, determination of its type, grading, and spread. This provides patients with an opportunity for early treatment interventions that can enhance treatment efficacy and improve survival rates. Timely and precisely diagnosis can help prevent tumor progression to advanced stages by reducing the complexity and difficulty associated with treatment while alleviating both physical and psychological burdens on patients [ 5 ] . Furthermore, accurate diagnosis aids doctors in developing personalized treatment plans encompassing surgery, chemotherapy, radiotherapy among other modalities to minimize patient discomfort while enhancing their overall quality of life [ 6 ] . Currently, commonly employed techniques for the diagnosis of CRC encompass colonoscopy, fecal occult blood examination, blood marker detection, and imaging examinations such as CT scanning [ 4 ] . Nevertheless, these methods possess inherent limitations [ 7 ] . For instance, colonoscopy necessitates patient cooperation and is not universally applicable due to its costliness; fecal occult blood examination may yield false positive or false negative outcomes; blood marker detection lacks specificity and can be influenced by other factors; imaging examinations fail to accurately differentiate between benign and malignant lesions [ 8 ] . Consequently, despite their certain utility in diagnosing CRC, these methods still exhibit constraints. It is imperative to comprehensively consider clinical manifestations, risk factors, and medical resources when selecting appropriate diagnostic approaches. In clinical practice, a comprehensive evaluation combined with clinical symptoms, colonoscopy findings, and histopathological examination results typically serves as the gold standard for diagnosing CRC [ 8 ] . The field of clinical diagnosis assisted by artificial intelligence is rapidly advancing [ 9 ] . Its primary advantage lies in leveraging machine learning and deep learning technologies to analyze vast amounts of medical data, including images, genomes, and clinical records, thereby aiding doctors in making quicker and more accurate decisions regarding diagnosis and treatment [ 10 ] . Artificial intelligence has demonstrated significant potential in tumor screening, disease classification, image recognition, and other domains [ 11 ] . It not only enhances the precision and efficiency of diagnoses but also fosters the development of personalized medicine by offering patients more precise and tailored treatment plans [ 12 ] . With ongoing technological advancements and increased integration into clinical practice, AI-assisted clinical diagnosis will bring about revolutionary changes in healthcare delivery while elevating the standard of medical care provided to enhance patients' quality of life [ 13 ] . In this study, we successfully identified diagnostic markers for CRC through bioinformatics analysis and the Random forest method. These markers were systematically analyzed using data mining (TCGA and GEO database) and multi-omics analysis (genomics and proteomics), resulting in the identification of a group of genes with significant differential expression that play crucial roles in the occurrence and development of CRC. This discovery provides new insights into diagnosis possibilities for CRC, offering strong support for potential applications in its diagnosis and treatment. To further enhance accuracy and sensitivity in clinical practice, we conducted machine learning (SVM)on immunohistochemical images of target proteins to construct a comprehensive diagnostic model. Through analyzing large amounts of immunohistochemical image data, we successfully established a diagnostic model that comprehensively considers different marker expression patterns, providing a more accurate auxiliary tool for diagnosing CRC. This breakthrough is expected to bring important progress to clinical practice. Results Differentially expressed genes in CRC We enrolled three datasets from the GEO database, which included both CRC and normal tissues. To enhance the credibility and stability of our findings, we initially merged these three datasets to increase the sample size in our study. Through differential analysis (Fig. 1 A and 1 B), we identified a total of 374 DEGs. Similarly, by performing the same analysis in the TCGA database, we obtained 2,514 DEGs encoding proteins (Fig. 1 C). By intersecting the results from both databases, we ultimately identified 232 genes with consistently altered expression in CRC (Fig. 1 D). Disease-Characteristic Genes of CRC In Fig. 2 A, the X-axis represents the number of trees, while the Y-axis represents the cross-validation error. The three curves in the figure represent errors for the control group, experimental group, and all samples respectively. The all-sample error is depicted by a black line. Our objective was to identify the point with minimal cross-validation error to determine the corresponding number of trees and genes. In Fig. 2 B, gene names are represented on the Y-axis, whereas gene importance scores are shown on the X-axis. A higher score indicates greater gene importance. We selected genes with a score exceeding 1 for further investigation, resulting in a total of 24 characteristic genes associated with CRC. Revalidation of protein levels When the aforementioned characteristic genes of CRC were inputted into the CPTAC database, we discovered that 6 proteins exhibited consistency with the results of genomics analysis and fulfilled the criteria for subsequent analysis. The disparities in protein expression between tumor and normal tissues, as well as across various tumor stages, are depicted in Fig. 3 A-L. Through validation in the HPA database, we ultimately confirmed that 4 proteins displaying significantly differential expression would be incorporated into the construction of machine learning models via IHC. The immunohistochemical maps illustrating these proteins in both normal tissues and bowel cancer tissues can be observed in Fig. 4 , while their quantitative outcomes are presented in Table 1 . Table 1 Expression of disease-associated proteins in normal and CRC tissues from HPA data. AKR1B10 CA2 DHRS9 ZG16 Normal Endocrine cells High Endothelial cells Not detected Not detected Medium Not detected Enterocytes High Enterocytes - Microvilli High Fibroblasts Not detected Glandular cells Medium Medium High Goblet cells High Mucosal lymphoid cells Not detected Peripheral nerve/ganglion Medium Not detected Cancer High Medium 2 1 Low 7 2 Not detected 10 12 3 8 CRC Machine Learning Model Based on Immunohistochemical Graph Construction The schematic diagram of the CRC diagnosis based on machine learning of immunohistochemical pathological images is presented in Fig. 5 A. The framework for dataset division, model training, and prediction evaluation is illustrated in Fig. 5 B. Additionally, Fig. 5 C displays the ROC curve and binary confusion matrix obtained using the immunohistochemical staining images from the HPA database. AKR1B10, CA2, DHRS9, and ZG16 were utilized as four biomarkers respectively. Among them, DHRS9 exhibited the highest accuracy (0.710) in diagnosing CRC compared to others such as ZG16 which had lower accuracy (0.550) and specificity (0.531), as shown in Table 2 . In order to enhance the sensitivity and specificity, we utilized the diagnostic markers and performed joint diagnosis using the SVM algorithm. Based on the ROC curve and binary confusion matrix depicted in Fig. 5 D, it is remarkably observed that the model accurately classified all 400 normal colon tissues, with only 5 out of 400 colon cancer tissues being misdiagnosed as normal tissues. This resulted in an exceptional accuracy and specificity of 0.999 (Table 2 , Supplementary 1). The proposed model introduces a novel approach for diagnosing CRC, showcasing the potential of machine learning in analyzing immunohistochemical pathological images while providing robust support for clinical diagnosis and practice. Table 2 Binary diagnostic performance for CRC based on immunohistochemical staining images. # of images AUROC Accuracy Precision Recall Specificity - + AKR1B10 100 100 0.648 (0.644–0.650) 0.600 (0.598–0.602) 0.520 (0.517–0.522) 0.645 (0.642–0.647) 0.532 (0.529–0.534) CA2 100 100 0.758 (0.755–0.760) 0.685 (0.682–0.687) 0.510 (0.506–0.514) 0.632 (0.628–0.636) 0.665 (0.663–0.667) DHRS9 100 100 0.865 (0.863–0.866) 0.710 (0.708–0.711) 0.540 (0.537–0.542) 0.899 (0.897-0.900) 0.676 (0.674–0.677) ZG16 100 100 0.598 (0.596–0.599) 0.550 (0.548–0.551) 0.310 (0.308–0.311) 0.639 (0.636–0.641) 0.531 (0.530–0.532) All 400 400 0.999 (0.999–0.999) 0.999 (0.999–0.999) 0.999 (0.999–0.999) 0.999 (0.999–0.999) 0.999 (0.999–0.999) (95%CI) Discussion The highlight of this study lies in the acquisition of a set of characteristic genes associated with CRC through multi-omics research and machine learning methods. By leveraging the features extracted from immunohistochemical images, we have developed a highly robust diagnostic model using machine learning techniques, which significantly contributes to the diagnosis of CRC and advances precision medicine. Multi-omics analysis refers to the comprehensive utilization of diverse biological data types including genomics, transcriptomics, proteomics, metabolomics, etc., aiming to gain a holistic understanding of the complexity and diversity within biological systems [ 13 ] . Its advantage lies in its ability to simultaneously consider multiple levels of biological information ranging from genes, RNA molecules, proteins to metabolites; thus enabling a more comprehensive and profound comprehension of tumor diseases [ 14 ] . In this study, potential characteristic markers for CRC were identified through comprehensive analysis and cross-validation across different datasets. These markers may be implicated in tumor initiation, progression as well as treatment response; thereby facilitating improved diagnosis and treatment outcomes for this disease. After conducting comprehensive validation of differential analysis (Fig. 1 A-D), employing Random Forest algorithm (Fig. 2 A, B), assessing protein expression levels (Fig. 3 A-L), and performing IHC analysis (Fig. 4 ), we ultimately identified four specific markers for CRC, namely AKR1B10, CA2, DHRS9, and ZG16. These aforementioned markers exhibited high expression in colon tissues but were either lowly expressed or not expressed at all in CRC. It is noteworthy that there is a paucity of literature on the four biomarkers. Under physiological conditions, AKR1B10 and DHRS9 belong to the aldosterone reductase family and ketone alcohol conversion SDR family respectively [ 15 , 16 ] ; they are involved in intracellular reactions associated with aldosterone reduction and ketone alcohol conversion processes which regulate the balance of intracellular metabolites. Previous research has demonstrated that apart from participating in the progression of CRC through their own metabolic reactions, AKR1B10 can also promote its advancement by influencing autophagy responses as well as inflammatory reactions [ 17 – 19 ] . CA2 is a widely expressed carbonic anhydrase isoenzyme in the human body, playing crucial roles in regulating acid-base balance, transporting carbon dioxide, maintaining calcium ion homeostasis, protecting the digestive tract and facilitating nervous system function [ 20 ] . Although the molecular biological mechanism of CA2 promoting CRC progression remains unknown, several studies have confirmed that its expression can significantly inhibit the cancer cell growth [ 21 ] ; furthermore, its downregulation represents an early event in CRC development [ 22 ] . ZG16 is a protein secreted by the pancreas involved in various biological processes such as pancreatic and digestive regulation, immune modulation and intestinal microbial regulation [ 23 ] . Meng H et al.'s research has demonstrated that ZG16 mainly contributes to CRC progression through tumor immune microenvironment [ 24 ] . In current clinical practice, IHC combined with the clinical manifestations of patients, imaging, serum markers, and colonoscopy serves as the gold standard for diagnosing CRC [ 25 , 26 ] . Its advantages include high specificity and the ability to provide comprehensive information on tissue structure and protein expression, facilitating a more thorough understanding of the pathological characteristics [ 27 ] . However, it is important to address and resolve technical complexities, specimen quality issues, and repeatability concerns associated with IHC [ 28 ] . Additionally, due to its semi-quantitative nature and susceptibility to subjective judgment by operators, there may be inherent subjectivity and uncertainty in immunohistochemical results [ 29 ] . Consequently, variations between different laboratories or operators can compromise diagnostic repeatability and consistency. Based on machine learning and deep learning technologies, artificial intelligence can process large-scale medical images and biological data, enabling doctors to extract potential lesions and abnormal areas from complex image data [ 30 ] . This provides a more comprehensive and objective diagnostic basis while reducing the interference of human factors [ 31 ] . Therefore, the application of artificial intelligence has brought new opportunities and challenges for tumor diagnosis, providing crucial support in improving medical standards and patient survival rates. Through the HPA database, we obtained immunohistochemical images of the aforementioned four biomarkers in CRC for machine learning-based image recognition and classification. The process diagram of the machine learning-based diagnosis model is illustrated in Fig. 5 A. It can be observed from the flow chart presented in Fig. 5 B that our classifier's performance was evaluated using support vector machine (SVM) and a 10-fold cross-validation approach. Furthermore, to demonstrate the superiority of joint judgment over individual biomarkers' judgments, we incorporated features extracted from other samples within the same class (cancer or normal) into the current image's feature set during joint judgment coding. The code primarily iterates through image files within a designated folder, reads each image, extracts both brown and Gabor filter features, and ultimately concatenates them to form the final feature vector. In each cycle of cross-validation, 9 subsets were selected as the training set while the remaining 1 subset was used as the test set. The performance of the classifier in cancer detection task was evaluated by calculating the confusion matrix, which included metrics such as accuracy, precision, and recall. This process was repeated for each biomarker to individually evaluate their classification performance. For the combined biomarker dataset, features were extracted from a single biomarker and then connected and input into 10 cross-validation processes to assess the impact of combining multiple biomarkers. Finally, the classification results obtained from a single biomarker were compared with those obtained from combined biomarkers to validate their superior joint judgment capability. The AUC area under ROC curve and diagnostic accuracy, specificity, recall, and predictive rate for each of the 4 individual biomarkers were obtained from Fig. 5 C and Table S1 in supporting materials. According to Fig. 5 D and comprehensive evaluation presented in Table 2 , it is evident that combining all four biomarkers yields significantly better diagnostic outcomes compared to independent diagnosis alone. Moreover, when considering accuracy, specificity, recall rate, predictive rate,and AUC area together,the combined diagnosis using all four biomarkers can achieve nearly perfect results (close to100%), which holds great significance for clinical diagnosis. The limitations of this study are as follows: Firstly, the data utilized in this study solely originate from the public data platform. To enhance the reliability and validity, future studies should consider incorporating clinical samples. Secondly, given the nature of machine learning and data analysis techniques, it is imperative to expand the sample size in order to obtain more precise and stable outcomes. Lastly, although this study is based on IHC findings without simplifying diagnostic efficiency or steps, there is potential for improvement by collaborating with colleagues of chemical materials to update detection procedures and refine the procedure. In summary, this study utilized multiple databases and conducted multi-omics analysis to identify four novel and promising protein markers (AKR1B10, CA2, DHRS9, and ZG16) for colorectal cancer. These markers were screened using bioinformatics and machine learning techniques (Random Forest). Furthermore, a CRC diagnosis model was constructed based on IHC employing another machine learning method (Support Vector Machine), achieving remarkable accuracy and specificity of 0.999. This research provides significant opportunities and support for enhancing the comprehensiveness and precision of CRC. Materials and methods Data Acquisition and Difference Analysis We selected three independent datasets (GSE18105, GSE21510, GSE33114) from the GEO database and merged them using the SVA package in R language to construct a GEO CRC cohort. Given that this study will be based on proteomics analysis and will build a machine learning model using immunohistochemical images, we extracted RNA-seq data of protein genes from the TCGA database to construct a TCGA CRC cohort. Subsequently, we utilized the limma package to analyze differences between the two CRC cohorts. Genes with p-values ≤ 0.05 and |logFC| ≥ 2 were identified as differentially expressed genes (DEGs) in CRC. Finally, we obtained the intersection of DEGs exhibiting stable differential expression in both databases for further studies. Refinement of disease-associated gene screening We utilized the random forest algorithm to filter DEGs and identify disease-related characteristic genes with specific parameter settings (seed = 123456 and tree = 500). By optimizing cross-validation error, we constructed an accurate random forest model. Through this screening process, we effectively identified significant characteristic genes that hold great potential for disease diagnosis, thereby providing crucial insights and directions for future research. The validation of proteomics We incorporated disease-associated genes of CRC into both the CPTAC and HPA databases for protein-level analysis. In the CPTAC database, we ensured significant differences in target proteins between tumor and normal tissues, while also examining differences among different tumor stages, particularly between normal tissues and Stage 1 (early colorectal cancer patients). Developing a Machine Learning Model based on CRC Immunohistochemistry (IHC) The Python programming language and multiple libraries are utilized for implementation. Initially, we import the following libraries: numpy for efficient processing of arrays and matrices, os for seamless manipulation of files and folders, cv2 for advanced image processing capabilities, sklearn for machine learning-related operations, and matplotlib for generating informative graphs. This combination of libraries offers robust functionality and flexibility, enabling us to effectively process data, perform intricate image processing tasks, implement sophisticated machine learning models, and visualize results. Consequently, we successfully develop an analysis and diagnosis model specifically tailored to identify characteristic genes associated with colorectal cancer. The code iterates through the image files within a designated folder. It reads each image file while extracting both brownian features as well as Gabor filter features. Brown channel features are assigned based on the brown color in the RGB color space. By creating a mask that identifies pixels falling within this brown range (set to white − 255), all other pixels are set to black (0). Subsequently, this mask is adjusted to a specified size before being normalized. Gabor filter features are obtained by applying the Gabor filter technique to grayscale images. Finally, these extracted brownian features along with Gabor filter features are concatenated together forming the final feature vector. Additionally,the corresponding labels of each image are stored in an array denoted as Y. During each cross-validation cycle,nine subsets from our dataset serve as training sets while one remaining subset functions as a test set. The classifier is then trained using the training set, and predictions are made on the test set. The performance of the classifier in cancer recognition task is evaluated by calculating the confusion matrix and various performance metrics, including accuracy, precision, recall, etc. This process enables us to assess the classifier's generalization ability across different datasets and better evaluate its practical performance. Declarations Author contributions Bobin Ning and Jimei Chi contributed to the design of the study and the acquisition and analysis of the data. Qingyu Meng drafted the work and Baoqing Jia revised it. Data availability statement The datasets generated during and/or analysed during the current study are available in the GEO dataset (https://www.ncbi.nlm.nih.gov/geo/) and the TCGA dataset (https://portal.gdc.cancer.gov/). Competing interests The authors declare no competing interests. References Dekker E, Tanis PJ, Vleugels J, Kasi PM, Wallace MB. Colorectal cancer. Lancet. 2019. 394(10207): 1467-1480. Baidoun F, Elshiwy K, Elkeraie Y, et al. Colorectal Cancer Epidemiology: Recent Trends and Impact on Outcomes. Curr Drug Targets. 2021. 22(9): 998-1009. Patel SG, Karlitz JJ, Yen T, Lieu CH, Boland CR. The rising tide of early-onset colorectal cancer: a comprehensive review of epidemiology, clinical features, biology, risk factors, prevention, and early detection. Lancet Gastroenterol Hepatol. 2022. 7(3): 262-274. Biller LH, Schrag D. Diagnosis and Treatment of Metastatic Colorectal Cancer: A Review. JAMA. 2021. 325(7): 669-685. Shin AE, Giancotti FG, Rustgi AK. Metastatic colorectal cancer: mechanisms and emerging therapeutics. Trends Pharmacol Sci. 2023. 44(4): 222-236. Mahmoud NN. Colorectal Cancer: Preoperative Evaluation and Staging. Surg Oncol Clin N Am. 2022. 31(2): 127-141. Heinimann K. [Hereditary Colorectal Cancer: Clinics, Diagnostics and Management]. Ther Umsch. 2018. 75(10): 601-606. Wu Z, Li Y, Zhang Y, et al. Colorectal Cancer Screening Methods and Molecular Markers for Early Detection. Technol Cancer Res Treat. 2020. 19: 1533033820980426. Sharma A, Kumar R, Yadav G, Garg P. Artificial intelligence in intestinal polyp and colorectal cancer prediction. Cancer Lett. 2023. 565: 216238. Mitsala A, Tsalikidis C, Pitiakoudis M, Simopoulos C, Tsaroucha AK. Artificial Intelligence in Colorectal Cancer Screening, Diagnosis and Treatment. A New Era. Curr Oncol. 2021. 28(3): 1581-1607. Foersch S, Glasner C, Woerl AC, et al. Multistain deep learning for prediction of prognosis and therapy response in colorectal cancer. Nat Med. 2023. 29(2): 430-439. Rompianesi G, Pegoraro F, Ceresa CD, Montalti R, Troisi RI. Artificial intelligence in the diagnosis and management of colorectal cancer liver metastases. World J Gastroenterol. 2022. 28(1): 108-122. Qiu H, Ding S, Liu J, Wang L, Wang X. Applications of Artificial Intelligence in Screening, Diagnosis, Treatment, and Prognosis of Colorectal Cancer. Curr Oncol. 2022. 29(3): 1773-1795. Fernandez-Rozadilla C, Timofeeva M, Chen Z, et al. Deciphering colorectal cancer genetics through multi-omic analysis of 100,204 cases and 154,587 controls of European and east Asian ancestries. Nat Genet. 2023. 55(1): 89-99. Salabei JK, Li XP, Petrash JM, Bhatnagar A, Barski OA. Functional expression of novel human and murine AKR1B genes. Chem Biol Interact. 2011. 191(1-3): 177-84. Napoli JL. Physiological insights into all-trans-retinoic acid biosynthesis. Biochim Biophys Acta. 2012. 1821(1): 152-67. Liu C, Shi L, Li W, et al. AKR1B10 accelerates the production of proinflammatory cytokines via the NF-κB signaling pathway in colon cancer. J Mol Histol. 2022. 53(5): 781-791. Li W, Liu C, Huang Z, et al. AKR1B10 negatively regulates autophagy through reducing GAPDH upon glucose starvation in colon cancer. J Cell Sci. 2021. 134(8). Shen Y, Ma J, Yan R, et al. Impaired self-renewal and increased colitis and dysplastic lesions in colonic mucosa of AKR1B8-deficient mice. Clin Cancer Res. 2015. 21(6): 1466-76. Viikilä P, Kivelä AJ, Mustonen H, et al. Carbonic anhydrase enzymes II, VII, IX and XII in colorectal carcinomas. World J Gastroenterol. 2016. 22(36): 8168-77. Nannini G, De Luca V, D', et al. A comparative study of carbonic anhydrase activity in lymphocytes from colorectal cancer tissues and adjacent healthy counterparts. J Enzyme Inhib Med Chem. 2022. 37(1): 1651-1655. Eldehna WM, Mohammed EE, Al-Ansary GH, et al. Design and synthesis of 6-arylpyridine-tethered sulfonamides as novel selective inhibitors of carbonic anhydrase IX with promising antitumor features toward the human colorectal cancer. Eur J Med Chem. 2023. 258: 115538. Meng H, Li W, Boardman LA, Wang L. Loss of ZG16 is associated with molecular and clinicopathological phenotypes of colorectal cancer. BMC Cancer. 2018. 18(1): 433. Meng H, Ding Y, Liu E, Li W, Wang L. ZG16 regulates PD-L1 expression and promotes local immunity in colon cancer. Transl Oncol. 2021. 14(2): 101003. Kryeziu K, Bergsland CH, Guren TK, Sveen A, Lothe RA. Multiplex immunohistochemistry of metastatic colorectal cancer and ex vivo tumor avatars. Biochim Biophys Acta Rev Cancer. 2022. 1877(1): 188682. Bărbălan A, Nicolaescu AC, Măgăran AV, et al. Immunohistochemistry predictive markers for primary colorectal cancer tumors: where are we and where are we going. Rom J Morphol Embryol. 2018. 59(1): 29-42. Sukswai N, Khoury JD. Immunohistochemistry Innovations for Diagnosis and Tissue-Based Biomarker Detection. Curr Hematol Malig Rep. 2019. 14(5): 368-375. Magaki S, Hojat SA, Wei B, So A, Yong WH. An Introduction to the Performance of Immunohistochemistry. Methods Mol Biol. 2019. 1897: 289-298. Choi JH, Ro JY. The 2020 WHO Classification of Tumors of Soft Tissue: Selected Changes and New Entities. Adv Anat Pathol. 2021. 28(1): 44-58. Swanson K, Wu E, Zhang A, Alizadeh AA, Zou J. From patterns to patients: Advances in clinical machine learning for cancer diagnosis, prognosis, and treatment. Cell. 2023. 186(8): 1772-1791. Tran KA, Kondrashova O, Bradley A, Williams ED, Pearson JV, Waddell N. Deep learning in cancer diagnosis, prognosis and treatment selection. Genome Med. 2021. 13(1): 152. Additional Declarations No competing interests reported. Supplementary Files Supplemantary1.docx Cite Share Download PDF Status: Published Journal Publication published 02 Dec, 2024 Read the published version in Scientific Reports → Version 1 posted Editorial decision: Revision requested 22 Jul, 2024 Reviews received at journal 15 Jul, 2024 Reviewers agreed at journal 12 Jul, 2024 Reviews received at journal 02 Jun, 2024 Reviewers agreed at journal 30 May, 2024 Reviewers agreed at journal 19 Apr, 2024 Reviewers agreed at journal 05 Apr, 2024 Reviewers invited by journal 05 Apr, 2024 Editor assigned by journal 05 Apr, 2024 Editor invited by journal 02 Apr, 2024 Submission checks completed at journal 02 Apr, 2024 First submitted to journal 19 Mar, 2024 You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-4129792","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Article","associatedPublications":[],"authors":[{"id":288516810,"identity":"645502f2-cb72-4a86-9837-e7bc0ed55b3a","order_by":0,"name":"Bobin Ning","email":"","orcid":"","institution":"Chinese PLA General Hospital","correspondingAuthor":false,"prefix":"","firstName":"Bobin","middleName":"","lastName":"Ning","suffix":""},{"id":288516812,"identity":"15e8a7b1-8aaa-438a-ae91-a85a339e6e67","order_by":1,"name":"Jimei Chi","email":"","orcid":"","institution":"Chinese Academy of Sciences (ICCAS), Beijing National Laboratory for Molecular Sciences (BNLMS)","correspondingAuthor":false,"prefix":"","firstName":"Jimei","middleName":"","lastName":"Chi","suffix":""},{"id":288516814,"identity":"aa71e724-2b0a-4478-a3cd-0890f23b9ff3","order_by":2,"name":"Qingyu Meng","email":"","orcid":"","institution":"Chinese PLA General Hospital","correspondingAuthor":false,"prefix":"","firstName":"Qingyu","middleName":"","lastName":"Meng","suffix":""},{"id":288516816,"identity":"ff042a74-9823-448c-aef7-b06df153668c","order_by":3,"name":"Baoqing Jia","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA00lEQVRIiWNgGAWjYBACAxCRwMDAzMbAfODAhwrStLAlHpxxhlgtEMBjfJi3hQgt5uyHD954uKOWnY/9zIcDvA0M8vxiB/BrsexJS7ZIPHOcmY0nd8MByR0MhjNnJxBw2IEcM4nEtmNAvwC1GJ5hSDC4TUjL+TdQLfxvHhxIbCNGyw2wLTXMbBI5DAcOEqPFcsYzoF/aDgC1PDM42HBGgrBfzPmTD9782VaXLN+f/PjznwobeX5pAlpAQIKB4XAygk0MACqrsyNO6SgYBaNgFIxIAAAIM0d0+gUptAAAAABJRU5ErkJggg==","orcid":"","institution":"Chinese PLA General Hospital","correspondingAuthor":true,"prefix":"","firstName":"Baoqing","middleName":"","lastName":"Jia","suffix":""}],"badges":[],"createdAt":"2024-03-19 11:12:44","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-4129792/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-4129792/v1","draftVersion":[],"editorialEvents":[{"content":"https://doi.org/10.1038/s41598-024-76083-9","type":"published","date":"2024-12-02T15:57:10+00:00"}],"editorialNote":"","failedWorkflow":false,"files":[{"id":54311883,"identity":"050aefdf-194b-4522-8e89-4e54d342b007","added_by":"auto","created_at":"2024-04-08 17:18:31","extension":"jpg","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":3277936,"visible":true,"origin":"","legend":"\u003cp\u003eDifferential analysis of genes encoding proteins in CRC. A. Heatmap of the most significantly differentially expressed 100 genes in GEO merged CRC cohort; B. Volcano plot of differentially expressed genes in GEO merged CRC cohort (|logFC| ≥2, p value≤0.05); C. Volcano plot of differentially expressed genes encoding proteins in TCGA CRC cohort (|logFC| ≥2, p value ≤0.05); D. Veen plot of DEGs encoding proteins of CRC patients in GEO and TCGA databases.\u003c/p\u003e","description":"","filename":"Figure1.jpg","url":"https://assets-eu.researchsquare.com/files/rs-4129792/v1/f360b0a383f2ba73d1156fc3.jpg"},{"id":54311882,"identity":"5271ba19-4fc2-4a3b-a7af-9772337eb79f","added_by":"auto","created_at":"2024-04-08 17:18:30","extension":"jpg","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":946505,"visible":true,"origin":"","legend":"\u003cp\u003eIdentification of characteristic genes of CRC by machine learning. A. The construction of Random Forest algorithmbased on DEGs; B. Identification of CRC characteristic genes based on significance scores.\u003c/p\u003e","description":"","filename":"Figure2.jpg","url":"https://assets-eu.researchsquare.com/files/rs-4129792/v1/c7d4c53aecfb73e67bcd8dff.jpg"},{"id":54311886,"identity":"f7909cad-56d0-4eac-bb4a-998078b01a8d","added_by":"auto","created_at":"2024-04-08 17:18:31","extension":"jpg","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":734164,"visible":true,"origin":"","legend":"\u003cp\u003eValidation of differential expression of disease-associated genes at protein level by CPTAC database. Expression differences between normal tissues and tumor tissues, A. AKR1B10, B. CA2, C. CEMIP, D. DHRS9,E. ITM2C, and F. ZG16; expression differences between normal tissues and each stage of tumor, A. AKR1B10, B. CA2, C. CEMIP, D. DHRS9,E. ITM2C, and F. ZG16.\u003c/p\u003e","description":"","filename":"Figure3.jpg","url":"https://assets-eu.researchsquare.com/files/rs-4129792/v1/05fff18f4aabda9558338135.jpg"},{"id":54311885,"identity":"2bf622d7-3ae7-4fc2-8ef3-24750de3b20c","added_by":"auto","created_at":"2024-04-08 17:18:31","extension":"jpg","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":3114661,"visible":true,"origin":"","legend":"\u003cp\u003eCharacteristic immunohistochemical images of AKR1B10, CA2, DHRS9, and ZG16 in HPA database.\u003c/p\u003e","description":"","filename":"Figure4.jpg","url":"https://assets-eu.researchsquare.com/files/rs-4129792/v1/9c6954b336ec037774cf00df.jpg"},{"id":54311887,"identity":"47d778fd-fd2b-4bef-9e00-8b8be6b05d55","added_by":"auto","created_at":"2024-04-08 17:18:32","extension":"jpg","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":11483263,"visible":true,"origin":"","legend":"\u003cp\u003ePathological diagnosis of CRC by machine learning. (A) Schematic diagram of machine learning-based immunohistochemical pathological images diagnosis of intestinal cancer. (B) Framework for partitioning data sets, training models and assessing predictions. (C) ROC curves and binary confusion matrices of AKR1B10, CA2, DHRS9 and ZG16 biomarkers based on immunohistochemical staining images taken by the HPA database, respectively. (D) ROC curves and binary confusion matrices of AKR1B10, CA2, DHRS9 and ZG16 combined diagnosis based on the SVM algorithm.\u003c/p\u003e","description":"","filename":"Figure5.jpg","url":"https://assets-eu.researchsquare.com/files/rs-4129792/v1/72c6f628681b2278e1054987.jpg"},{"id":70964664,"identity":"0ffd577e-fb4c-404d-be54-00a25779b97e","added_by":"auto","created_at":"2024-12-09 16:13:55","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":20092197,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-4129792/v1/91ba0cc2-cf7d-4217-bcbe-ffd8ebf98749.pdf"},{"id":54311884,"identity":"a6b12409-aede-4732-a277-9e1a2223d779","added_by":"auto","created_at":"2024-04-08 17:18:31","extension":"docx","order_by":1,"title":"","display":"","copyAsset":false,"role":"supplement","size":19470,"visible":true,"origin":"","legend":"","description":"","filename":"Supplemantary1.docx","url":"https://assets-eu.researchsquare.com/files/rs-4129792/v1/2e9bab28b0c3690069e9ce76.docx"}],"financialInterests":"No competing interests reported.","formattedTitle":"Accurate prediction of colorectal cancer diagnosis using machine learning based on immunohistochemistry pathological images","fulltext":[{"header":"Introduction","content":"\u003cp\u003eColorectal cancer (CRC) is a prevalent malignant tumor, and its incidence is progressively increasing due to the changing unhealthy lifestyle habits, dietary patterns, and aging\u003csup\u003e[\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e]\u003c/sup\u003e. It primarily affects individuals over 50 years old\u003csup\u003e[\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e]\u003c/sup\u003e. Despite advancements in early screening and treatment methods that have improved patient survival rates, CRC remains one of the leading causes of death worldwide\u003csup\u003e[\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e]\u003c/sup\u003e. Many patients are diagnosed at an advanced stage, which significantly increases the risk of recurrence and metastasis\u003csup\u003e[\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e]\u003c/sup\u003e. Therefore, timely diagnosis plays a crucial role for CRC patients as it enables early detection of the disease's presence, determination of its type, grading, and spread. This provides patients with an opportunity for early treatment interventions that can enhance treatment efficacy and improve survival rates. Timely and precisely diagnosis can help prevent tumor progression to advanced stages by reducing the complexity and difficulty associated with treatment while alleviating both physical and psychological burdens on patients\u003csup\u003e[\u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e]\u003c/sup\u003e. Furthermore, accurate diagnosis aids doctors in developing personalized treatment plans encompassing surgery, chemotherapy, radiotherapy among other modalities to minimize patient discomfort while enhancing their overall quality of life\u003csup\u003e[\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e]\u003c/sup\u003e.\u003c/p\u003e \u003cp\u003eCurrently, commonly employed techniques for the diagnosis of CRC encompass colonoscopy, fecal occult blood examination, blood marker detection, and imaging examinations such as CT scanning\u003csup\u003e[\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e]\u003c/sup\u003e. Nevertheless, these methods possess inherent limitations\u003csup\u003e[\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e]\u003c/sup\u003e. For instance, colonoscopy necessitates patient cooperation and is not universally applicable due to its costliness; fecal occult blood examination may yield false positive or false negative outcomes; blood marker detection lacks specificity and can be influenced by other factors; imaging examinations fail to accurately differentiate between benign and malignant lesions\u003csup\u003e[\u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e]\u003c/sup\u003e. Consequently, despite their certain utility in diagnosing CRC, these methods still exhibit constraints. It is imperative to comprehensively consider clinical manifestations, risk factors, and medical resources when selecting appropriate diagnostic approaches. In clinical practice, a comprehensive evaluation combined with clinical symptoms, colonoscopy findings, and histopathological examination results typically serves as the gold standard for diagnosing CRC\u003csup\u003e[\u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e]\u003c/sup\u003e.\u003c/p\u003e \u003cp\u003eThe field of clinical diagnosis assisted by artificial intelligence is rapidly advancing\u003csup\u003e[\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e]\u003c/sup\u003e. Its primary advantage lies in leveraging machine learning and deep learning technologies to analyze vast amounts of medical data, including images, genomes, and clinical records, thereby aiding doctors in making quicker and more accurate decisions regarding diagnosis and treatment\u003csup\u003e[\u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e]\u003c/sup\u003e. Artificial intelligence has demonstrated significant potential in tumor screening, disease classification, image recognition, and other domains\u003csup\u003e[\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e]\u003c/sup\u003e. It not only enhances the precision and efficiency of diagnoses but also fosters the development of personalized medicine by offering patients more precise and tailored treatment plans\u003csup\u003e[\u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e]\u003c/sup\u003e. With ongoing technological advancements and increased integration into clinical practice, AI-assisted clinical diagnosis will bring about revolutionary changes in healthcare delivery while elevating the standard of medical care provided to enhance patients' quality of life\u003csup\u003e[\u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e]\u003c/sup\u003e.\u003c/p\u003e \u003cp\u003eIn this study, we successfully identified diagnostic markers for CRC through bioinformatics analysis and the Random forest method. These markers were systematically analyzed using data mining (TCGA and GEO database) and multi-omics analysis (genomics and proteomics), resulting in the identification of a group of genes with significant differential expression that play crucial roles in the occurrence and development of CRC. This discovery provides new insights into diagnosis possibilities for CRC, offering strong support for potential applications in its diagnosis and treatment. To further enhance accuracy and sensitivity in clinical practice, we conducted machine learning (SVM)on immunohistochemical images of target proteins to construct a comprehensive diagnostic model. Through analyzing large amounts of immunohistochemical image data, we successfully established a diagnostic model that comprehensively considers different marker expression patterns, providing a more accurate auxiliary tool for diagnosing CRC. This breakthrough is expected to bring important progress to clinical practice.\u003c/p\u003e"},{"header":"Results","content":"\u003cdiv id=\"Sec3\" class=\"Section2\"\u003e \u003ch2\u003eDifferentially expressed genes in CRC\u003c/h2\u003e \u003cp\u003eWe enrolled three datasets from the GEO database, which included both CRC and normal tissues. To enhance the credibility and stability of our findings, we initially merged these three datasets to increase the sample size in our study. Through differential analysis (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003eA and \u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003eB), we identified a total of 374 DEGs. Similarly, by performing the same analysis in the TCGA database, we obtained 2,514 DEGs encoding proteins (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003eC). By intersecting the results from both databases, we ultimately identified 232 genes with consistently altered expression in CRC (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003eD).\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec4\" class=\"Section2\"\u003e \u003ch2\u003eDisease-Characteristic Genes of CRC\u003c/h2\u003e \u003cp\u003eIn Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003eA, the X-axis represents the number of trees, while the Y-axis represents the cross-validation error. The three curves in the figure represent errors for the control group, experimental group, and all samples respectively. The all-sample error is depicted by a black line. Our objective was to identify the point with minimal cross-validation error to determine the corresponding number of trees and genes. In Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003eB, gene names are represented on the Y-axis, whereas gene importance scores are shown on the X-axis. A higher score indicates greater gene importance. We selected genes with a score exceeding 1 for further investigation, resulting in a total of 24 characteristic genes associated with CRC.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec5\" class=\"Section2\"\u003e \u003ch2\u003eRevalidation of protein levels\u003c/h2\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eWhen the aforementioned characteristic genes of CRC were inputted into the CPTAC database, we discovered that 6 proteins exhibited consistency with the results of genomics analysis and fulfilled the criteria for subsequent analysis. The disparities in protein expression between tumor and normal tissues, as well as across various tumor stages, are depicted in Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003eA-L. Through validation in the HPA database, we ultimately confirmed that 4 proteins displaying significantly differential expression would be incorporated into the construction of machine learning models via IHC. The immunohistochemical maps illustrating these proteins in both normal tissues and bowel cancer tissues can be observed in Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003e, while their quantitative outcomes are presented in Table\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab1\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 1\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eExpression of disease-associated proteins in normal and CRC tissues from HPA data.\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"6\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c6\" colnum=\"6\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003eAKR1B10\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003eCA2\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c5\"\u003e \u003cp\u003eDHRS9\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c6\"\u003e \u003cp\u003eZG16\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eNormal\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eEndocrine cells\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eHigh\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eEndothelial cells\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eNot detected\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eNot detected\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eMedium\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003eNot detected\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eEnterocytes\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eHigh\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eEnterocytes - Microvilli\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eHigh\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eFibroblasts\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eNot detected\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eGlandular cells\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eMedium\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eMedium\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003eHigh\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eGoblet cells\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eHigh\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eMucosal lymphoid cells\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eNot detected\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003ePeripheral nerve/ganglion\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eMedium\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003eNot detected\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eCancer\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eHigh\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eMedium\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e2\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e1\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eLow\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e7\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e2\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eNot detected\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e10\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e12\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e3\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e8\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec6\" class=\"Section2\"\u003e \u003ch2\u003eCRC Machine Learning Model Based on Immunohistochemical Graph Construction\u003c/h2\u003e \u003cp\u003eThe schematic diagram of the CRC diagnosis based on machine learning of immunohistochemical pathological images is presented in Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003eA. The framework for dataset division, model training, and prediction evaluation is illustrated in Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003eB. Additionally, Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003eC displays the ROC curve and binary confusion matrix obtained using the immunohistochemical staining images from the HPA database. AKR1B10, CA2, DHRS9, and ZG16 were utilized as four biomarkers respectively. Among them, DHRS9 exhibited the highest accuracy (0.710) in diagnosing CRC compared to others such as ZG16 which had lower accuracy (0.550) and specificity (0.531), as shown in Table\u0026nbsp;\u003cspan refid=\"Tab2\" class=\"InternalRef\"\u003e2\u003c/span\u003e. In order to enhance the sensitivity and specificity, we utilized the diagnostic markers and performed joint diagnosis using the SVM algorithm. Based on the ROC curve and binary confusion matrix depicted in Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003eD, it is remarkably observed that the model accurately classified all 400 normal colon tissues, with only 5 out of 400 colon cancer tissues being misdiagnosed as normal tissues. This resulted in an exceptional accuracy and specificity of 0.999 (Table\u0026nbsp;\u003cspan refid=\"Tab2\" class=\"InternalRef\"\u003e2\u003c/span\u003e, Supplementary 1). The proposed model introduces a novel approach for diagnosing CRC, showcasing the potential of machine learning in analyzing immunohistochemical pathological images while providing robust support for clinical diagnosis and practice.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab2\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 2\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eBinary diagnostic performance for CRC based on immunohistochemical staining images.\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"8\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c6\" colnum=\"6\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c7\" colnum=\"7\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c8\" colnum=\"8\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\" morerows=\"1\" rowspan=\"2\"\u003e\u0026nbsp;\u003c/th\u003e \u003cth align=\"left\" colspan=\"2\" nameend=\"c3\" namest=\"c2\"\u003e \u003cp\u003e# of images\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\" morerows=\"1\" rowspan=\"2\"\u003e \u003cp\u003eAUROC\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c5\" morerows=\"1\" rowspan=\"2\"\u003e \u003cp\u003eAccuracy\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c6\" morerows=\"1\" rowspan=\"2\"\u003e \u003cp\u003ePrecision\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c7\" morerows=\"1\" rowspan=\"2\"\u003e \u003cp\u003eRecall\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c8\" morerows=\"1\" rowspan=\"2\"\u003e \u003cp\u003eSpecificity\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003e\u003cb\u003e-\u003c/b\u003e\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003e\u003cb\u003e+\u003c/b\u003e\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eAKR1B10\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e100\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e100\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.648\u003c/p\u003e \u003cp\u003e(0.644\u0026ndash;0.650)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.600\u003c/p\u003e \u003cp\u003e(0.598\u0026ndash;0.602)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e0.520\u003c/p\u003e \u003cp\u003e(0.517\u0026ndash;0.522)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e0.645\u003c/p\u003e \u003cp\u003e(0.642\u0026ndash;0.647)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e0.532\u003c/p\u003e \u003cp\u003e(0.529\u0026ndash;0.534)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eCA2\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e100\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e100\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.758\u003c/p\u003e \u003cp\u003e(0.755\u0026ndash;0.760)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.685\u003c/p\u003e \u003cp\u003e(0.682\u0026ndash;0.687)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e0.510\u003c/p\u003e \u003cp\u003e(0.506\u0026ndash;0.514)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e0.632\u003c/p\u003e \u003cp\u003e(0.628\u0026ndash;0.636)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e0.665\u003c/p\u003e \u003cp\u003e(0.663\u0026ndash;0.667)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eDHRS9\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e100\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e100\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.865\u003c/p\u003e \u003cp\u003e(0.863\u0026ndash;0.866)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.710\u003c/p\u003e \u003cp\u003e(0.708\u0026ndash;0.711)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e0.540\u003c/p\u003e \u003cp\u003e(0.537\u0026ndash;0.542)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e0.899\u003c/p\u003e \u003cp\u003e(0.897-0.900)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e0.676\u003c/p\u003e \u003cp\u003e(0.674\u0026ndash;0.677)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eZG16\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e100\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e100\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.598\u003c/p\u003e \u003cp\u003e(0.596\u0026ndash;0.599)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.550\u003c/p\u003e \u003cp\u003e(0.548\u0026ndash;0.551)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e0.310\u003c/p\u003e \u003cp\u003e(0.308\u0026ndash;0.311)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e0.639\u003c/p\u003e \u003cp\u003e(0.636\u0026ndash;0.641)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e0.531\u003c/p\u003e \u003cp\u003e(0.530\u0026ndash;0.532)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eAll\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e400\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e400\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.999\u003c/p\u003e \u003cp\u003e(0.999\u0026ndash;0.999)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.999\u003c/p\u003e \u003cp\u003e(0.999\u0026ndash;0.999)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e0.999\u003c/p\u003e \u003cp\u003e(0.999\u0026ndash;0.999)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e0.999\u003c/p\u003e \u003cp\u003e(0.999\u0026ndash;0.999)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e0.999\u003c/p\u003e \u003cp\u003e(0.999\u0026ndash;0.999)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colspan=\"8\" nameend=\"c8\" namest=\"c1\"\u003e \u003cp\u003e(95%CI)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003c/div\u003e"},{"header":"Discussion","content":"\u003cp\u003eThe highlight of this study lies in the acquisition of a set of characteristic genes associated with CRC through multi-omics research and machine learning methods. By leveraging the features extracted from immunohistochemical images, we have developed a highly robust diagnostic model using machine learning techniques, which significantly contributes to the diagnosis of CRC and advances precision medicine. Multi-omics analysis refers to the comprehensive utilization of diverse biological data types including genomics, transcriptomics, proteomics, metabolomics, etc., aiming to gain a holistic understanding of the complexity and diversity within biological systems\u003csup\u003e[\u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e]\u003c/sup\u003e. Its advantage lies in its ability to simultaneously consider multiple levels of biological information ranging from genes, RNA molecules, proteins to metabolites; thus enabling a more comprehensive and profound comprehension of tumor diseases\u003csup\u003e[\u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e]\u003c/sup\u003e. In this study, potential characteristic markers for CRC were identified through comprehensive analysis and cross-validation across different datasets. These markers may be implicated in tumor initiation, progression as well as treatment response; thereby facilitating improved diagnosis and treatment outcomes for this disease.\u003c/p\u003e \u003cp\u003eAfter conducting comprehensive validation of differential analysis (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003eA-D), employing Random Forest algorithm (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003eA, B), assessing protein expression levels (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003eA-L), and performing IHC analysis (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003e), we ultimately identified four specific markers for CRC, namely AKR1B10, CA2, DHRS9, and ZG16. These aforementioned markers exhibited high expression in colon tissues but were either lowly expressed or not expressed at all in CRC. It is noteworthy that there is a paucity of literature on the four biomarkers. Under physiological conditions, AKR1B10 and DHRS9 belong to the aldosterone reductase family and ketone alcohol conversion SDR family respectively\u003csup\u003e[\u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e, \u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e16\u003c/span\u003e]\u003c/sup\u003e; they are involved in intracellular reactions associated with aldosterone reduction and ketone alcohol conversion processes which regulate the balance of intracellular metabolites. Previous research has demonstrated that apart from participating in the progression of CRC through their own metabolic reactions, AKR1B10 can also promote its advancement by influencing autophagy responses as well as inflammatory reactions\u003csup\u003e[\u003cspan additionalcitationids=\"CR18\" citationid=\"CR17\" class=\"CitationRef\"\u003e17\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e19\u003c/span\u003e]\u003c/sup\u003e. CA2 is a widely expressed carbonic anhydrase isoenzyme in the human body, playing crucial roles in regulating acid-base balance, transporting carbon dioxide, maintaining calcium ion homeostasis, protecting the digestive tract and facilitating nervous system function\u003csup\u003e[\u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e20\u003c/span\u003e]\u003c/sup\u003e. Although the molecular biological mechanism of CA2 promoting CRC progression remains unknown, several studies have confirmed that its expression can significantly inhibit the cancer cell growth\u003csup\u003e[\u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e21\u003c/span\u003e]\u003c/sup\u003e; furthermore, its downregulation represents an early event in CRC development\u003csup\u003e[\u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e22\u003c/span\u003e]\u003c/sup\u003e. ZG16 is a protein secreted by the pancreas involved in various biological processes such as pancreatic and digestive regulation, immune modulation and intestinal microbial regulation\u003csup\u003e[\u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e23\u003c/span\u003e]\u003c/sup\u003e. Meng H et al.'s research has demonstrated that ZG16 mainly contributes to CRC progression through tumor immune microenvironment\u003csup\u003e[\u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e24\u003c/span\u003e]\u003c/sup\u003e.\u003c/p\u003e \u003cp\u003eIn current clinical practice, IHC combined with the clinical manifestations of patients, imaging, serum markers, and colonoscopy serves as the gold standard for diagnosing CRC\u003csup\u003e[\u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e25\u003c/span\u003e, \u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e26\u003c/span\u003e]\u003c/sup\u003e. Its advantages include high specificity and the ability to provide comprehensive information on tissue structure and protein expression, facilitating a more thorough understanding of the pathological characteristics\u003csup\u003e[\u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e27\u003c/span\u003e]\u003c/sup\u003e. However, it is important to address and resolve technical complexities, specimen quality issues, and repeatability concerns associated with IHC\u003csup\u003e[\u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e28\u003c/span\u003e]\u003c/sup\u003e. Additionally, due to its semi-quantitative nature and susceptibility to subjective judgment by operators, there may be inherent subjectivity and uncertainty in immunohistochemical results\u003csup\u003e[\u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e29\u003c/span\u003e]\u003c/sup\u003e. Consequently, variations between different laboratories or operators can compromise diagnostic repeatability and consistency.\u003c/p\u003e \u003cp\u003eBased on machine learning and deep learning technologies, artificial intelligence can process large-scale medical images and biological data, enabling doctors to extract potential lesions and abnormal areas from complex image data\u003csup\u003e[\u003cspan citationid=\"CR30\" class=\"CitationRef\"\u003e30\u003c/span\u003e]\u003c/sup\u003e. This provides a more comprehensive and objective diagnostic basis while reducing the interference of human factors\u003csup\u003e[\u003cspan citationid=\"CR31\" class=\"CitationRef\"\u003e31\u003c/span\u003e]\u003c/sup\u003e. Therefore, the application of artificial intelligence has brought new opportunities and challenges for tumor diagnosis, providing crucial support in improving medical standards and patient survival rates. Through the HPA database, we obtained immunohistochemical images of the aforementioned four biomarkers in CRC for machine learning-based image recognition and classification. The process diagram of the machine learning-based diagnosis model is illustrated in Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003eA. It can be observed from the flow chart presented in Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003eB that our classifier's performance was evaluated using support vector machine (SVM) and a 10-fold cross-validation approach. Furthermore, to demonstrate the superiority of joint judgment over individual biomarkers' judgments, we incorporated features extracted from other samples within the same class (cancer or normal) into the current image's feature set during joint judgment coding. The code primarily iterates through image files within a designated folder, reads each image, extracts both brown and Gabor filter features, and ultimately concatenates them to form the final feature vector. In each cycle of cross-validation, 9 subsets were selected as the training set while the remaining 1 subset was used as the test set. The performance of the classifier in cancer detection task was evaluated by calculating the confusion matrix, which included metrics such as accuracy, precision, and recall. This process was repeated for each biomarker to individually evaluate their classification performance. For the combined biomarker dataset, features were extracted from a single biomarker and then connected and input into 10 cross-validation processes to assess the impact of combining multiple biomarkers. Finally, the classification results obtained from a single biomarker were compared with those obtained from combined biomarkers to validate their superior joint judgment capability. The AUC area under ROC curve and diagnostic accuracy, specificity, recall, and predictive rate for each of the 4 individual biomarkers were obtained from Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003eC and Table \u003cspan refid=\"MOESM1\" class=\"InternalRef\"\u003eS1\u003c/span\u003e in supporting materials. According to Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003eD and comprehensive evaluation presented in Table\u0026nbsp;\u003cspan refid=\"Tab2\" class=\"InternalRef\"\u003e2\u003c/span\u003e, it is evident that combining all four biomarkers yields significantly better diagnostic outcomes compared to independent diagnosis alone. Moreover, when considering accuracy, specificity, recall rate, predictive rate,and AUC area together,the combined diagnosis using all four biomarkers can achieve nearly perfect results (close to100%), which holds great significance for clinical diagnosis.\u003c/p\u003e \u003cp\u003eThe limitations of this study are as follows: Firstly, the data utilized in this study solely originate from the public data platform. To enhance the reliability and validity, future studies should consider incorporating clinical samples. Secondly, given the nature of machine learning and data analysis techniques, it is imperative to expand the sample size in order to obtain more precise and stable outcomes. Lastly, although this study is based on IHC findings without simplifying diagnostic efficiency or steps, there is potential for improvement by collaborating with colleagues of chemical materials to update detection procedures and refine the procedure.\u003c/p\u003e \u003cp\u003eIn summary, this study utilized multiple databases and conducted multi-omics analysis to identify four novel and promising protein markers (AKR1B10, CA2, DHRS9, and ZG16) for colorectal cancer. These markers were screened using bioinformatics and machine learning techniques (Random Forest). Furthermore, a CRC diagnosis model was constructed based on IHC employing another machine learning method (Support Vector Machine), achieving remarkable accuracy and specificity of 0.999. This research provides significant opportunities and support for enhancing the comprehensiveness and precision of CRC.\u003c/p\u003e"},{"header":"Materials and methods","content":"\u003cdiv id=\"Sec9\" class=\"Section2\"\u003e \u003ch2\u003eData Acquisition and Difference Analysis\u003c/h2\u003e \u003cp\u003eWe selected three independent datasets (GSE18105, GSE21510, GSE33114) from the GEO database and merged them using the SVA package in R language to construct a GEO CRC cohort. Given that this study will be based on proteomics analysis and will build a machine learning model using immunohistochemical images, we extracted RNA-seq data of protein genes from the TCGA database to construct a TCGA CRC cohort. Subsequently, we utilized the limma package to analyze differences between the two CRC cohorts. Genes with p-values\u0026thinsp;\u0026le;\u0026thinsp;0.05 and |logFC| \u0026ge; 2 were identified as differentially expressed genes (DEGs) in CRC. Finally, we obtained the intersection of DEGs exhibiting stable differential expression in both databases for further studies.\u003c/p\u003e \u003cdiv id=\"Sec10\" class=\"Section3\"\u003e \u003ch2\u003eRefinement of disease-associated gene screening\u003c/h2\u003e \u003cp\u003eWe utilized the random forest algorithm to filter DEGs and identify disease-related characteristic genes with specific parameter settings (seed\u0026thinsp;=\u0026thinsp;123456 and tree\u0026thinsp;=\u0026thinsp;500). By optimizing cross-validation error, we constructed an accurate random forest model. Through this screening process, we effectively identified significant characteristic genes that hold great potential for disease diagnosis, thereby providing crucial insights and directions for future research.\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv id=\"Sec11\" class=\"Section2\"\u003e \u003ch2\u003eThe validation of proteomics\u003c/h2\u003e \u003cp\u003eWe incorporated disease-associated genes of CRC into both the CPTAC and HPA databases for protein-level analysis. In the CPTAC database, we ensured significant differences in target proteins between tumor and normal tissues, while also examining differences among different tumor stages, particularly between normal tissues and Stage 1 (early colorectal cancer patients).\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec12\" class=\"Section2\"\u003e \u003ch2\u003eDeveloping a Machine Learning Model based on CRC Immunohistochemistry (IHC)\u003c/h2\u003e \u003cp\u003eThe Python programming language and multiple libraries are utilized for implementation. Initially, we import the following libraries: numpy for efficient processing of arrays and matrices, os for seamless manipulation of files and folders, cv2 for advanced image processing capabilities, sklearn for machine learning-related operations, and matplotlib for generating informative graphs. This combination of libraries offers robust functionality and flexibility, enabling us to effectively process data, perform intricate image processing tasks, implement sophisticated machine learning models, and visualize results. Consequently, we successfully develop an analysis and diagnosis model specifically tailored to identify characteristic genes associated with colorectal cancer. The code iterates through the image files within a designated folder. It reads each image file while extracting both brownian features as well as Gabor filter features. Brown channel features are assigned based on the brown color in the RGB color space. By creating a mask that identifies pixels falling within this brown range (set to white \u0026minus;\u0026thinsp;255), all other pixels are set to black (0). Subsequently, this mask is adjusted to a specified size before being normalized. Gabor filter features are obtained by applying the Gabor filter technique to grayscale images. Finally, these extracted brownian features along with Gabor filter features are concatenated together forming the final feature vector. Additionally,the corresponding labels of each image are stored in an array denoted as Y. During each cross-validation cycle,nine subsets from our dataset serve as training sets while one remaining subset functions as a test set. The classifier is then trained using the training set, and predictions are made on the test set. The performance of the classifier in cancer recognition task is evaluated by calculating the confusion matrix and various performance metrics, including accuracy, precision, recall, etc. This process enables us to assess the classifier's generalization ability across different datasets and better evaluate its practical performance.\u003c/p\u003e \u003c/div\u003e"},{"header":"Declarations","content":"\u003cp\u003e\u003cstrong\u003eAuthor contributions\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eBobin Ning and Jimei Chi contributed to the design of the study and the acquisition and analysis of the data. Qingyu Meng drafted the work and Baoqing Jia revised it.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eData availability statement\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe datasets generated during and/or analysed during the current study are available in the GEO dataset (https://www.ncbi.nlm.nih.gov/geo/) and the TCGA dataset (https://portal.gdc.cancer.gov/).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eCompeting interests\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe authors declare no competing interests.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\n\u003cli\u003eDekker E, Tanis PJ, Vleugels J, Kasi PM, Wallace MB. Colorectal cancer. Lancet. 2019. 394(10207): 1467-1480.\u003c/li\u003e\n\u003cli\u003eBaidoun F, Elshiwy K, Elkeraie Y, et al. Colorectal Cancer Epidemiology: Recent Trends and Impact on Outcomes. Curr Drug Targets. 2021. 22(9): 998-1009.\u003c/li\u003e\n\u003cli\u003ePatel SG, Karlitz JJ, Yen T, Lieu CH, Boland CR. The rising tide of early-onset colorectal cancer: a comprehensive review of epidemiology, clinical features, biology, risk factors, prevention, and early detection. Lancet Gastroenterol Hepatol. 2022. 7(3): 262-274.\u003c/li\u003e\n\u003cli\u003eBiller LH, Schrag D. Diagnosis and Treatment of Metastatic Colorectal Cancer: A Review. JAMA. 2021. 325(7): 669-685.\u003c/li\u003e\n\u003cli\u003eShin AE, Giancotti FG, Rustgi AK. Metastatic colorectal cancer: mechanisms and emerging therapeutics. Trends Pharmacol Sci. 2023. 44(4): 222-236.\u003c/li\u003e\n\u003cli\u003eMahmoud NN. Colorectal Cancer: Preoperative Evaluation and Staging. Surg Oncol Clin N Am. 2022. 31(2): 127-141.\u003c/li\u003e\n\u003cli\u003eHeinimann K. [Hereditary Colorectal Cancer: Clinics, Diagnostics and Management]. Ther Umsch. 2018. 75(10): 601-606.\u003c/li\u003e\n\u003cli\u003eWu Z, Li Y, Zhang Y, et al. Colorectal Cancer Screening Methods and Molecular Markers for Early Detection. Technol Cancer Res Treat. 2020. 19: 1533033820980426.\u003c/li\u003e\n\u003cli\u003eSharma A, Kumar R, Yadav G, Garg P. Artificial intelligence in intestinal polyp and colorectal cancer prediction. Cancer Lett. 2023. 565: 216238.\u003c/li\u003e\n\u003cli\u003eMitsala A, Tsalikidis C, Pitiakoudis M, Simopoulos C, Tsaroucha AK. Artificial Intelligence in Colorectal Cancer Screening, Diagnosis and Treatment. A New Era. Curr Oncol. 2021. 28(3): 1581-1607.\u003c/li\u003e\n\u003cli\u003eFoersch S, Glasner C, Woerl AC, et al. Multistain deep learning for prediction of prognosis and therapy response in colorectal cancer. Nat Med. 2023. 29(2): 430-439.\u003c/li\u003e\n\u003cli\u003eRompianesi G, Pegoraro F, Ceresa CD, Montalti R, Troisi RI. Artificial intelligence in the diagnosis and management of colorectal cancer liver metastases. World J Gastroenterol. 2022. 28(1): 108-122.\u003c/li\u003e\n\u003cli\u003eQiu H, Ding S, Liu J, Wang L, Wang X. Applications of Artificial Intelligence in Screening, Diagnosis, Treatment, and Prognosis of Colorectal Cancer. Curr Oncol. 2022. 29(3): 1773-1795.\u003c/li\u003e\n\u003cli\u003eFernandez-Rozadilla C, Timofeeva M, Chen Z, et al. Deciphering colorectal cancer genetics through multi-omic analysis of 100,204 cases and 154,587 controls of European and east Asian ancestries. Nat Genet. 2023. 55(1): 89-99.\u003c/li\u003e\n\u003cli\u003eSalabei JK, Li XP, Petrash JM, Bhatnagar A, Barski OA. Functional expression of novel human and murine AKR1B genes. Chem Biol Interact. 2011. 191(1-3): 177-84.\u003c/li\u003e\n\u003cli\u003eNapoli JL. Physiological insights into all-trans-retinoic acid biosynthesis. Biochim Biophys Acta. 2012. 1821(1): 152-67.\u003c/li\u003e\n\u003cli\u003eLiu C, Shi L, Li W, et al. AKR1B10 accelerates the production of proinflammatory cytokines via the NF-\u0026kappa;B signaling pathway in colon cancer. J Mol Histol. 2022. 53(5): 781-791.\u003c/li\u003e\n\u003cli\u003eLi W, Liu C, Huang Z, et al. AKR1B10 negatively regulates autophagy through reducing GAPDH upon glucose starvation in colon cancer. J Cell Sci. 2021. 134(8).\u003c/li\u003e\n\u003cli\u003eShen Y, Ma J, Yan R, et al. Impaired self-renewal and increased colitis and dysplastic lesions in colonic mucosa of AKR1B8-deficient mice. Clin Cancer Res. 2015. 21(6): 1466-76.\u003c/li\u003e\n\u003cli\u003eViikil\u0026auml; P, Kivel\u0026auml; AJ, Mustonen H, et al. Carbonic anhydrase enzymes II, VII, IX and XII in colorectal carcinomas. World J Gastroenterol. 2016. 22(36): 8168-77.\u003c/li\u003e\n\u003cli\u003eNannini G, De Luca V, D\u0026amp;#x27, et al. A comparative study of carbonic anhydrase activity in lymphocytes from colorectal cancer tissues and adjacent healthy counterparts. J Enzyme Inhib Med Chem. 2022. 37(1): 1651-1655.\u003c/li\u003e\n\u003cli\u003eEldehna WM, Mohammed EE, Al-Ansary GH, et al. Design and synthesis of 6-arylpyridine-tethered sulfonamides as novel selective inhibitors of carbonic anhydrase IX with promising antitumor features toward the human colorectal cancer. Eur J Med Chem. 2023. 258: 115538.\u003c/li\u003e\n\u003cli\u003eMeng H, Li W, Boardman LA, Wang L. Loss of ZG16 is associated with molecular and clinicopathological phenotypes of colorectal cancer. BMC Cancer. 2018. 18(1): 433.\u003c/li\u003e\n\u003cli\u003eMeng H, Ding Y, Liu E, Li W, Wang L. ZG16 regulates PD-L1 expression and promotes local immunity in colon cancer. Transl Oncol. 2021. 14(2): 101003.\u003c/li\u003e\n\u003cli\u003eKryeziu K, Bergsland CH, Guren TK, Sveen A, Lothe RA. Multiplex immunohistochemistry of metastatic colorectal cancer and ex vivo tumor avatars. Biochim Biophys Acta Rev Cancer. 2022. 1877(1): 188682.\u003c/li\u003e\n\u003cli\u003eBărbălan A, Nicolaescu AC, Măgăran AV, et al. Immunohistochemistry predictive markers for primary colorectal cancer tumors: where are we and where are we going. Rom J Morphol Embryol. 2018. 59(1): 29-42.\u003c/li\u003e\n\u003cli\u003eSukswai N, Khoury JD. Immunohistochemistry Innovations for Diagnosis and Tissue-Based Biomarker Detection. Curr Hematol Malig Rep. 2019. 14(5): 368-375.\u003c/li\u003e\n\u003cli\u003eMagaki S, Hojat SA, Wei B, So A, Yong WH. An Introduction to the Performance of Immunohistochemistry. Methods Mol Biol. 2019. 1897: 289-298.\u003c/li\u003e\n\u003cli\u003eChoi JH, Ro JY. The 2020 WHO Classification of Tumors of Soft Tissue: Selected Changes and New Entities. Adv Anat Pathol. 2021. 28(1): 44-58.\u003c/li\u003e\n\u003cli\u003eSwanson K, Wu E, Zhang A, Alizadeh AA, Zou J. From patterns to patients: Advances in clinical machine learning for cancer diagnosis, prognosis, and treatment. Cell. 2023. 186(8): 1772-1791.\u003c/li\u003e\n\u003cli\u003eTran KA, Kondrashova O, Bradley A, Williams ED, Pearson JV, Waddell N. Deep learning in cancer diagnosis, prognosis and treatment selection. Genome Med. 2021. 13(1): 152.\u003c/li\u003e\n\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":false,"highlight":"","institution":"","isAcceptedByJournal":true,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"scientific-reports","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"scirep","sideBox":"Learn more about [Scientific Reports](http://www.nature.com/srep/)","snPcode":"","submissionUrl":"","title":"Scientific Reports","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"stoa","reportingPortfolio":"Scientific Reports","inReviewEnabled":true,"inReviewRevisionsEnabled":true},"keywords":"colorectal cancer, diagnosis, machine learning, immunohistochemistry ","lastPublishedDoi":"10.21203/rs.3.rs-4129792/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-4129792/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eColorectal cancer (CRC) ranks as the third most prevalent tumor and the second leading cause of mortality. Early and accurate diagnosis holds significant importance in enhancing patient treatment and prognosis. Machine learning technology and bioinformatics have provided novel approaches for cancer diagnosis. This study aims to develop a CRC diagnostic model based on immunohistochemical staining image features using machine learning methods. Initially, CRC disease-specific genes were identified through bioinformatics analysis and Random Forest algorithm utilizing RNA-seq data from both GEO and TCGA databases. Subsequently, verification of these genes was performed using proteomics data from CPTAC and HPA database, resulting in identification of target proteins (AKR1B10, CA2, DHRS9, and ZG16) for further investigation. SVM algorithm was then employed to analyze and integrate the characteristics of immunohistochemical images to construct a reliable CRC diagnostic model. During the training and validation process of this model, cross-validation along with external validation methods were implemented to ensure accuracy and reliability. The results demonstrate that the established diagnostic model exhibits excellent performance in distinguishing between CRC and normal controls (accuracy rate: 0.999), thereby presenting potential prospects for clinical application. These findings are expected to provide innovative perspectives as well as methodologies for personalized diagnosis of CRC while offering more precise references for promising treatment.\u003c/p\u003e","manuscriptTitle":"Accurate prediction of colorectal cancer diagnosis using machine learning based on immunohistochemistry pathological images","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2024-04-08 17:18:25","doi":"10.21203/rs.3.rs-4129792/v1","editorialEvents":[{"type":"communityComments","content":0},{"type":"decision","content":"Revision requested","date":"2024-07-22T18:22:45+00:00","index":"","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2024-07-15T11:31:32+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"191398132978625865636057534060186993808","date":"2024-07-13T00:07:47+00:00","index":"hide","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2024-06-02T11:39:53+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"309325941732438668562905985618830875413","date":"2024-05-31T01:57:42+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"b632340b-fd1d-484c-ac60-69579df3e80c","date":"2024-04-19T14:27:17+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"1735c0b2-d222-4e2e-8a4c-a87639cc1cc4","date":"2024-04-06T02:19:00+00:00","index":"hide","fulltext":""},{"type":"reviewersInvited","content":"","date":"2024-04-05T19:39:43+00:00","index":"","fulltext":""},{"type":"editorAssigned","content":"","date":"2024-04-05T19:38:30+00:00","index":"","fulltext":""},{"type":"editorInvited","content":"","date":"2024-04-02T18:05:20+00:00","index":"","fulltext":""},{"type":"checksComplete","content":"","date":"2024-04-02T18:04:25+00:00","index":"","fulltext":""},{"type":"submitted","content":"Scientific Reports","date":"2024-03-19T11:11:28+00:00","index":"","fulltext":""}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"scientific-reports","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"scirep","sideBox":"Learn more about [Scientific Reports](http://www.nature.com/srep/)","snPcode":"","submissionUrl":"","title":"Scientific Reports","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"stoa","reportingPortfolio":"Scientific Reports","inReviewEnabled":true,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"b742a818-e285-430b-bd2e-08c9e0a82a74","owner":[],"postedDate":"April 8th, 2024","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"published-in-journal","subjectAreas":[{"id":30374605,"name":"Biological sciences/Cancer"},{"id":30374606,"name":"Biological sciences/Computational biology and bioinformatics"},{"id":30374607,"name":"Health sciences/Biomarkers"},{"id":30374608,"name":"Health sciences/Gastroenterology"},{"id":30374609,"name":"Health sciences/Oncology"}],"tags":[],"updatedAt":"2024-12-09T16:00:15+00:00","versionOfRecord":{"articleIdentity":"rs-4129792","link":"https://doi.org/10.1038/s41598-024-76083-9","journal":{"identity":"scientific-reports","isVorOnly":false,"title":"Scientific Reports"},"publishedOn":"2024-12-02 15:57:10","publishedOnDateReadable":"December 2nd, 2024"},"versionCreatedAt":"2024-04-08 17:18:25","video":"","vorDoi":"10.1038/s41598-024-76083-9","vorDoiUrl":"https://doi.org/10.1038/s41598-024-76083-9","workflowStages":[]},"version":"v1","identity":"rs-4129792","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-4129792","identity":"rs-4129792","version":["v1"]},"buildId":"qtupq5eGEP_6zYnWcrvyt","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}