Analytical Validation of a Circulating Tumor DNA Assay using PhasED-Seq Technology for Detecting Residual Disease in B-Cell Malignancies

MRD , ctDN A, PhasED-Seq, CL ARITY , r esidual dis ease

The author s w ould l ik e t o acknowledge K ry st al Br own, PhD and St ep hanie Meek, PhD , f or their assist ance in prepara ti on o f this manuscript. All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprintthis version posted August 10, 2024. ; https://doi.org/10.1101/2024.08.09.24311742doi: medRxiv preprint NOTE: This preprint reports new research that has not been certified by peer review and should not be used to guide clinical practice. 2

Background : Cir cula ting tumo r DNA (ctDNA) is a non-in v asiv e biomark er th a t c an b e used as a tool t o de tect minimal r esidual dis ease (MRD). MRD c an pr ovide import a n t prognos t ic inf orma tion in dif f use lar ge B-cell lymphomas (DLBCL). Her e, w e present a n MRD assa y with an impr ov ed detection method f or ctDNA, Phased V arian t Enrichment and Det ec tio n Sequencing (PhasED-Seq) which lev er a g es phased v ari ants (PV s ) t o det ec t ctDNA .

Plasma samples fr om non-c ancer c o n t r ol s w er e used t o ass ess assa y specificity . A limiting dilution se ries using DLBC L clinic al-c on tr iv ed samples was perf ormed t o assess assa y sensitivity a nd pr ecision . The accur acy of the PhasED-Seq-based assa y w as assessed using plasma samples fr om individuals with DLBCL and f or whom MRD c ompar at or assa y re sults w ere also a v ailabl e. All sampl es w ere sour ced fr om c ommer cial v en dor s or a c ademic studies .

The analyti c al and clinical perf ormance o f the MRD assa y w as ev alua ted using clinic al and clinic al- c ontriv ed DLBCL samples. The assa y ’ s f alse positiv e rat e w as 0.24% and t he background error r ate w as 1.95E-08. The limit of det ec tion a t 95% det ec tion r ate (LoD95) a t 120 ng w as 0.7 p arts in 1,000,000 and pr ecision w as >96%. Clinical accur acy w as 90.62% PP A and 77.78% NP A.

The PhasED-Seq-based MRD assa y has s tr ong analyti c al and clinical perf ormanc e in B-cell dy scr asias. Thr ough the developme n t of impr ov ed ct DNA detection met hods such as th a t pre sent ed he r e , patie n t out comes ma y be impr ov ed th r ough th e det ec tion of r esid ual diseas e or ea rly r ela pse which ma y be used t o guid e treatme n t decisio ns. All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprintthis version posted August 10, 2024. ; https://doi.org/10.1101/2024.08.09.24311742doi: medRxiv preprint 3 Brief Summary Her e we present th e analy tic al v alid a ti on of a non-in v asiv e minimal r esidual dise a se (MRD) assa y which uses Phased V arian t Enrichment and Detection Seq uencing (PhasED-Seq) t o impr ov e the e rr or p r ofile and sensitivity of cir culating tumor DN A (ctDNA) det ec tion. Th e assa y ’ s perf orman ce included a f alse positiv e rat e of 0.24% and a backgr ound err o r r at e of 1.95E-08. The limit of de t ect ion a t 120 ng w as 0.7 parts in 1,000,000 (6.61E-07 PV AF) with pr ecision >96%. P ositive and nega tiv e ag reeme n t w e r e 90.62% and 77.78%, r esp ectiv ely . This sug g ests t ha t the PhasED-Seq-based MRD assa y is accur at e and r eproducible, thus appropri at e f or clinic a l use f or individuals with B-cell malignancies. All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprintthis version posted August 10, 2024. ; https://doi.org/10.1101/2024.08.09.24311742doi: medRxiv preprint 4

Cir cula ting tumo r DNA (ctDNA), tumor D NA shed into the bl oodstr e am, is a non-i n v asiv e biomark er that c an b e used as a t ool t o d et ec t minimal r esidu al disease (MRD). Det ec tion of c ance r- specific soma tic muta tions fr om ctDN A can pr ovide clinic ally relev ant inf orma tion t o predict t her ap eutic r espons e, diseas e r ecu rr e nce, and sur viv al, and thus guide i nt er v e n ti on decisions (1-3). As the utility of ctDNA detectio n has bec ome appreciate d, multiple in v e s tig ation al and c omme r ci ally-a v ailable methods of det ecti on ha v e b een developed (4). H ow ev er , th e sensitivity of fir s t-genera ti on appr oaches is limi t ed and impr ov ed metho ds ar e n eed ed to de t ec t r esidu al ctDNA when the tumo r burden is low and the individual has a higher pr o bability of responding t o ther a peut ic int er v e n tio n. Dif fuse lar g e B-cell lymphoma (DLBCL ) is the mos t c ommon type of non-Hodgkin l ymphoma (NHL) in the Unit ed S tat es (5). Despit e a tt empts t o increase the e f fic acy of c on v e ntional fir st-line immunochemotherapy ov er the p as t tw o dec ades , appr oximat ely 40% of DLBCL p a ti ents s till f ail to r espond o r r el apse (6). Curr ent DLBCL r esponse cri t eri a r ely on function al r adiogr a phic indices such as positr on emission t omograph y/ c ompute d t omogr aph y (PET /C T) sc ans, which ha v e limit ed s ensitivity and specificity (7). Ther e is a clear need t o de v elop precision tools c apa ble of r apid an d accur at e identific ation of patie n ts ha rboring resid ual c ancer bu r de n who ma y be a t high ris k of r elapse, such as the de t ecti on of r esidu al tumor in the bl ood (i.e. ctDN A-MRD). W e ha v e develope d a ctDNA-MRD pla tf orm base d on Phased V ariant Enrich men t and De t ect ion Sequ encing (PhasED -Seq) t o lev er a g e phased v aria n ts (PV s) t o impr ov e the sen sitivity of ctDNA detection (8). PV s ar e multiple soma tic mut ations in close pr oximity th a t c an be c oncurrently obser v ed on individual DNA molecules. PV s occur in mos t c ance r types bu t are pr ev ale n t in s t e r eo typed regions in B-cell malignanci es (8), and ar e an a ttr active tar g e t t o impr ov e molecul ar de t ec tion techniqu es giv en thei r intrinsic all y low err or pr ofile (9). Her e we describ e the a nalyti c al v alid a tio n of a sensitiv e PhasED-Seq-based MRD assa y .

MRD Assay Over view The F or esig h t CLARITY MRD assa y (F or esi gh t Diagnos tics, I nc.) w as assessed. This assa y utiliz es a tumor DNA sample (pre-tre a tme n t plasm a or tumor tissue), a non-c anc er ous o r normal DNA sample [e. g., periph er al blo od mononucle ar cell (PMBC) gDNA], and an MRD monit oring s ample (plasma). Extr ac t ed DN A fr om all samples is seque nced using a fixed hybrid captu re pan el (~150kb) tha t enriches for genomic regions in are as tha t recur re ntly harbor PVs in B-cell lymphomas. F oll owing sequencing , PV s All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprintthis version posted August 10, 2024. ; https://doi.org/10.1101/2024.08.09.24311742doi: medRxiv preprint 5 ar e id entified in the tumor and n on-c anc er ous DNA samples t o generat e a tumor- specific soma tic PV lis t. Tumor-specific PVs ar e de fin ed as thos e t ha t a r e p r ese n t in the tumo r DNA sample and absen t in the normal DNA sample . This tumor-specific PV lis t is then used t o ass ess f or MRD in the MRD monit o ring sample using inf orma tiv e molecul es, or a n y cell-fr ee DNA (cfDNA) molecules spanning the loca ti on of a tumor-specific PV tha t c ould ha rbor a PV (Figur e 1). Mut a n t mol ecules are inf ormativ e molecul es c ont aining th e mutant allel e. MRD is de fi ned as the p r ese nce of tumor specific PV s (mut an t molecul es), meeting a th r eshold b ased on th e lik elih ood of an incident al mu t ation ov e rlappin g with the tumor- specific PV lis t. Sample Prepar ati on The analyti c al perf orma nce of the PhasE D-Seq-based MRD assa y w as assessed in the Clinic al Labor at o ry Impr ov eme n t Amendme n ts (CLIA)-r egis t e r ed la bora tory at F o r esight Diagnos tics, Inc., f ollowing s t andar d op era ting pr oc edures. Samples included hea lth y donor sampl e s (self-r eported c ancer-fr e e at time of c oll ectio n; N =169), clinic al DLBCL samples (N=76), and clinic al-c ontriv ed DLBCL samples (N=2). Clinic al DLBCL samples w er e sampl es obt ain ed fr om individuals with an active diagnosis of DLBC L. Clinic al-c on t riv ed DLBCL samples w ere pr ep ared by c ombining e x tr ac t e d cell-fr ee DNA (cfDNA) fr om clinic al samples and he alth y donor samples; multipl e clinic al DLBCL and health y donor samples w er e p ooled t o mak e clinical-c o n trived s amples. The clinic al-contriv e d samples wer e then dilu t e d t o a t arg e t ed phas ed v aria n t all ele fr actio n (P V AF). PV AF is de fined as the ra tio of mole cules c ont aining a tumor-specific PV per molecules spannin g the positions of ≥1 PV . Both clinical and clinic al-c ontriv e d samples w ere r eq uir e d t o meet th e f ollowing crit eria f or study inclusion: minimu m input mass of 5 ng and ≥85% Phr ed quali ty sc or e of 30 (Q30) fr om the Illumina sequenc er . S equencin g metrics ar e repor ted within each s tu dy . Both c omme r cially pr ocu r ed (Disc ov ery Lif e Sciences and Bi oIVT) and r esidual samples fr om ac ademic resea r ch c ollab or ations w e r e u tiliz ed in th e analyti c al v alid a tio n s tudi es. Samples fr om ac ademic c oll abora tio ns w ere c oll ected with appr opr iat e p atient c ons ent which allow ed f or r ese ar ch use of r esidual samples and in s ti tutio nal r evi ew boar d (IRB) ov er sight. P ositiv e and n e g a tiv e contr ols w e r e used along with each study sample bat ch . The positiv e c ontr ol w as a mi x of lymphoma cell lines rich in PV s and the neg a tive c o n trol c onsist e d of libr aries prepa r ed with 50 µL nuclease-free w ater and carri ed thr ough th e e n ti r e w orkflow . All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprintthis version posted August 10, 2024. ; https://doi.org/10.1101/2024.08.09.24311742doi: medRxiv preprint 6 DNA Isola tion cfDNA w as isola ted fr om plasma using th e QIAs ymphon y DSP Cir cula ting DNA Kit (Qiag en, Hilden, G erman y; Cat alog N umber: 9375 56) on the automat ed Q IAs ymphon y s y s tem. Double-s tranded (dsDNA) w as quan tified by fluor ometry u sing a Qubit Fluor ome t e r with th e Qubit dsDNA High Sensitivity Assa y Kit (In vitr og en, W alth am, MA; C a ta log Number: Q32854). gDNA w as isol a te d fr om plasma- deple t ed whole blo od (PDWB) or PBMCs using the c ommer ci al QIAs ymphon y DSP DNA Mini Kit (Qiag en, Hilden, G erman y; Cat alog N umber: 9372 36) on the automat ed Q IAs ymphon y s y s tem and sheared using sonic ation . DNA w as qua n tified using th e Qubit dsDNA Br o ad Range Assa y Kit (In vitr ogen, W al tham, MA; Ca t al og Number : Q32853). Library Prepar ati on a nd Next- Gener ati on Sequenci ng Libr ary pr ep ar ation, h ybrid captu r e t arg e t enrichme n t , and sequ encing by s yn thesis w as perf ormed acc o r ding t o F o r esight Diagnos tics, Inc., optimiz ed w orkflow s unde r s ta ndar d op er ating pr ocedu r es. Five to 120 ng of cf DNA or gDNA w ere used t o c on s truc t seque ncing libr aries using KAP A HyperPr ep Kits (R oche Sequ encing Soluti ons, Indian apolis, I N) on manual and au tomat ed cust om w orkflow s. Libr ary DNA w as en riched usi ng a cus t om B-cell lymphoma pr ob e pan el (Int egr ated DNA T echnologies, Inc.), pe rf ormed per the manuf actur e r ’ s instructio ns using both a manual and an aut omat ed w o rkflow on the Br a v o Au t o ma ted Liquid Handling pl a tf orm. F ollowi ng enrichment, libr ar ies w er e s equenc ed using sequencing by s yn thesis on th e Illumina Nov aSeq 6000 and / or Nov aSeq X Plus ins trume n t (Illumin a, San Diego , CA). Anal ysis of Sequenci ng Da t a a nd MRD Status Det er mina tio n Sequence d at a w e r e an aly z ed using in-ho use dev elop ed algorithms and pip elines . Brie fly , r aw sequencing dat a were demultipl ex ed t o F AST Q files f or each sample using BCL Con v ert softw a r e (Illumina, San Diego , CA; V er sions 2.2.0 to 2.4.0). Low-quality sequencing reads a nd adapter read- thr ough w e r e r emov ed using f as tp (v er si on 0.20.0). Sequ encing r e ads w ere th en aligned to the re f e r ence g enome (G R Ch37) using BW A-MEM align er (v er sion 2.2 .1) t o cr e at e o ne alignme nt file per sample , f ollow ed by pr opriet a ry methods t o remov e polymer ase chain reacti on (PCR) and optic al duplicat es . The r esulti ng sequence alignment file w as used f or the anal y sis of PV s. MRD s t a tus w a s det ermin ed by the pr esenc e or absenc e of tumor-specific P V s, meeting a th r eshold bas ed on th e lik e lihood of an incident al mut ation ov e rlapping with th e tumo r-specific PV lis t. All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprintthis version posted August 10, 2024. ; https://doi.org/10.1101/2024.08.09.24311742doi: medRxiv preprint 7 Anal ytic al Specifici ty The assa y specificity or limit of blank (Lo B) w as ev alua ted acc o r ding to CLSI guidance EP17-A2 (10). EP17 -A2 de fines the LoB as t he high es t v alue e xp ected to be obser v e d fr om a series of measur em ents on a sample th at c ont ai ns no analyte (blank samples). Whole bl oo d fr om 60 c ancer-fr ee donor s (blank samples) w as c ollec t ed in Str eck cfDNA blood collecti on tub es (BCT s; Str eck, Cat al og Number : 230470), pr ocessed t o plasma, and cfDNA. DNA inpu t mass int o libr a ry pr epara t ion w as 120 ng. T w o libr ary replicat es w e r e p r ep ar e d fr o m each donor and 120 libr a ries w e r e gen er at ed f or sequ encing. Libr aries w e r e interr ogat ed by DLBCL tu mor-specific PV lis ts r esulting in a MRD p ositiv e or n eg a t iv e c all . The f alse positiv e ra te (FPR) and backgr ound err o r r at e f or t he assa y w as c alcul a te d per donor and ov er all. Anal ytic al Sensitivi ty T o det ermine the limit of de t ec tion (LoD) of the MRD assa y (95% det ection r ate pe r CLSI EP17- A2), a limit ed dilu tion seri es of a DLBCL cl inic al-c o n trived sample w as p r epa r ed a t 6 t ar geted PV AF lev els (7.00E-06, 3.50E-06, 1.75E-06, 8.75E-07, 4.38E-07, and 2.19E-07). Clinic al-c on t riv e d sample r epli c ates w er e c r e at ed by c ombining cfDNA fr om 4 DLBCL pa tie n t samples a nd diluting th e mixtu r e int o backgr ound cfDNA fr om health y dono r plasma. T en r epli c at es were test ed across 2 r eagent lots. Corr ec t ed t arg e t ed PV AF lev els, based on the obser v ed PV AF s, w e r e used f or de t e ction r at e and pr obit mod els. The de t ec tion r ate f or eac h lev el w as c alcul a ted by adding the number of MRD positiv e c alls and dividing the numbe r by the t ota l number of r epli c at es t e s ted. The p r obit model w as used t o c ompute th e number of mutant molecul es and PV AF c orr esp onding to a de t ecti o n r a te of 95% f or the sample. Pr obit mod el fit w as acceptable by ev alua ting with a s tatis ti c al goodness of fit t est. R eprod uci bilit y an d R epe atabili ty Assa y pr ecision w as ev alu a ted with a clin ic al-c ontrived MRD positive sample th a t w as pr epa r ed acr oss dif f erent tar g e ted PV AF lev els. A t 120 ng input mass, t arg e t ed PV AF lev els w er e 7 .00E-06, 3.50E- 06, and 1.75E-06 and a t 5 ng input mass t arg e t ed PV AF lev els w e r e 0.0001, 0.000 04, and 0.00002. A v er age positive agr e ement (AP A) w as used t o c alcul a te th e assa y ’ s r ep eat abil ity a nd r eproducibility as described in Y u et al , 2016. (11). Sample r eplicat es wer e p r ep ared acr oss 2 operator s, 2 r ea g e n t lo ts, and 3 time points. All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprintthis version posted August 10, 2024. ; https://doi.org/10.1101/2024.08.09.24311742doi: medRxiv preprint 8 Accura cy The accur acy of the PhasED-Seq-based MRD assa y w as det ermined by c ompari ng t he r esul ts to a pr eviously es tablished single nucl eotid e v ariant (SNV)-based orthogonal meth od f or detection of ctDN A using samples fr om individuals with DLBCL (12 ). Samples fr om a t otal of 19 individuals with DLB CL w er e utiliz ed, including 19 pre-treatme n t plas ma samples, 19 normal samples, 31 time poin t-of-int e r e s t plasma samples fr om timepoi n ts of int e res t duri ng tr e atment (Cy cle 2, Da y 1 or C y cle 3, Da y 1) or a t end- of-ther apy (E O T). R esults fr om the S NV-based MRD assa y and PhasED-Seq-based MRD assa y w er e c ompared. Disc ordant results bet w een t he tw o assa y s w e r e a djudic ated by c omparison with clinic al out comes.

Quality Co ntrol (QC) P ass Rat e The assa y QC pass r a t e across all pr e-an al ytic, analytic , and post-analytic met rics during the c onduct of this an alytical v alidation w as 99.0%. Anal ytic al Specifici ty The assa y specificity w as assessed using cfDNA fr om 60 c ancer-fr ee don or s (blank samples). All sample r epli c ates passed QC metrics f or 120 libr aries f or ev aluation. S amples w e re sequence d t o an a v er age median dep th of 21,321x and on -t ar g e t c ov e r ag e w as >91%. As blank samples do not ha ve tumor-derived PV s, PV lis ts fr om 35 DLBC L pa tie n ts w e r e us ed to measure th e FPR and backgr ound er r or r at e. The 35 PV lis ts c ov ered 65.5% of th e t o t al B-cell captu r e pan el spanning mul tiple chr omosomes. E ach blank sample r epli c ate w as int er r ogat ed by 35 pa ti ent PV lis ts r esulting in 4,2 00 possible tumor det ec tion c alls t o ev alu at e t he assa y FPR (T able 1). The ov er all assa y FPR w as 0.24%. The backgr ound err o r r at e of th e PhasED-Seq-based MRD assa y w as 1.95E-08, or 1.95 mut ant mol ecules in 100 million inf orma tiv e molecul es. Anal ytic al Sensitivi ty DLBCL clinic al-c on t riv ed sampl e r epli ca te s w er e p r epa r ed a t 6 t arg e t ed PV AF lev el s. All r epli c ates passed quality c o n t r ol met rics. R eplicat es w er e seq uenced t o a n a v er a g e median d epth of 19,455x. The PV lis t g ener at e d f or the clinic al-contriv e d sample had a total of 9,043 PV s. In a dilution seri es r anging fr om 4.83E-06 t o 1.51E-07, PV AF w as linear with the dilu tion of mutant molecules (Supplement a ry Figur e 1). Det ec tion r ates at each PV AF lev el ar e p r es ent ed in T ab le 2. Bas ed on probit modeling th e All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprintthis version posted August 10, 2024. ; https://doi.org/10.1101/2024.08.09.24311742doi: medRxiv preprint 9 mut ant molecul es and PV AF c orrespondi ng t o a de t ecti on r at e of 95%, the LoD of the MRD assa y , is 3.11 mut ant molecul es and 6.61E-07 (Supplemen tary Figure 2), or 0.7 par ts in 1,000,0 00. R eprod uci bilit y an d R epe atabili ty T able 3 show s the assa y precision using clinic al-c o n trived samples, c overing 5 ng a nd 120 ng DNA input mass and th e low t o high analy tic al measur eme n t rang e . All sample r eplicates acr oss operat or s, r eagent lots, and time poi n ts (N =104) passed QC metrics. R eplicat es w e r e se quen ced t o an a ver age median depth of 21,965x . Assa y r ep e a ta bility and r ep r oducibili ty w as >96% (T able 3). Accura cy Fifty samples w er e ev alu at ed f or c o nc o r d ance between th e PhasED-Seq-based MR D assa y and a pr eviously es tablished S NV-based metho d f or MRD det ection (12). All samples pa ssed QC metrics. Libr ary input mass f or plasma cfDNA samples f or MRD det ecti on r anged fr om 21.3 t o 80 ng and while non-c ancerous, normal sampl es w ere prepared at 80 ng. Normal sampl es w e r e s equenced t o a n a v er a g e median depth of 4012x an d cfDNA samples 5980x. The numbe r of phased v aria n ts f or each sample and donor w as c alcul at ed a nd r anged fr om 1 t o 1,816 PV s. Acc or ding t o th e c omparat o r SNV-based method, 18 samples w e r e called MRD ne g a tiv e an d 32 w er e c alled MRD positiv e (T able 4). MRD monit o ring samples th a t wer e c alled MR D positiv e by the c ompar ator assa y had a r ange of tumor f r actions fr om 0.000022 to 0.1697. Using the PhasED-Seq-based MRD assa y , 17 samples w er e c all ed MRD neg a tive and 33 samples MRD positive. T he MRD positiv e samples as determined by the PhasED-Se q-based MRD assa y had a r ang e of PV AF s fr om 0.0000088 t o 0.2567. P ositiv e per ce n t ag r eement (PP A) f or the MRD assa y w as 90.62% (95 % CI 74.98%, 98.02%) and NP A w as 77.78% (95% CI 52.73, 93.59; T able 4) using the SNV-based method as re f er enc e. The r e w e r e 7 disc or da n t c alls between the two metho ds. In all th e c ases whe r e t he two assa y s w er e dis c or d ant (T able 5) the PhasED-Seq-based MRD assa y agr eed with th e clinic al ou t comes (PP A 100%, NP A 100%). In the same c ases, the SNV-based c omparat or a ssa y had a low er c oncor da nce with clinical out comes (PP A 0%, NP A 60%).

Her e we describ e the a nalyti c al v alid a tio n of an MRD assa y tha t utilizes PhasED-S eq technology t o ide n tify tumo r-specific PV s t o det ec t ctDNA in B-cell dy scr asias. The mai n ben e fit of using PV s f or MRD All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprintthis version posted August 10, 2024. ; https://doi.org/10.1101/2024.08.09.24311742doi: medRxiv preprint 10 det ec tion is impr oving the signal- t o-nois e r a ti o in sequencing d a ta by r equiring t h e c onc o r da n t de t ec tion of a t lea s t 2 separat e non- re f e r ence events in an individual DNA molecule . Lev er ag ing multiple somatic mut ations within individual cfDNA fr agmen ts t o de t ec t ctDNA r educes the background error r ate, as pr eviously described (9). The use of multiple v ariants on th e same DNA str and (i.e . PV s) pr ovides an adv ant age at low ctDNA lev els in which t he specificity of SNV-based t echn ologies is r educed du e to the inherent backgr ound er r or rat e of SNV s. The analy ses prese n ted he r e d emons trate the an alytical perf ormanc e of this PhasED-Seq-based MRD assa y . The high specificity of the MRD assa y w as demons tr at e d through anal y sis of samples fr om individuals without cancer (N=60), with a FPR of 0.24% and a backgr ound err or rate of 1.95E-08, which is ~1000-f old low er than repor ted f or SNP-based technologies , ev en when u tilizing unique molecula r identifier s (9). With this low backgr ound err o r r at e , the a nalyti c al sensit ivity of the PhasED-Seq-based MRD assa y w as det ermine d t o b e 0.7 par t per 1 million (6.61E-07 PV AF). It is import a n t t o no te th a t th is analytical sensitivity w as de t e rmined in t he c onte x t of a limit e d amount of DNA in put. T o achiev e this sensitivity , an adequ at e pl asma sample p r oviding >1,000,000 inf orma tiv e mol ecules mus t be utilized; there f ore, st ar ting fr om a lar ger amou n t of plasma or cfDNA is r ec ommend ed. If a n a v ailable sample does not cont ain suf ficie n t cfDNA, the nu mber of inf orma tiv e molecul es will dict ate the level at which ctDNA c an b e detected. Inde ed, ev en the LoD s tudy described here is af f ec t ed by t he number of inf orma tiv e molecul es and DNA inpu t an d, with higher amou n ts of DNA inpu t, an impr ov ed LoD w ould be e xp ected. T ak en t o g eth er , th e assa y ’ s high sensitivit y and specificity sug g es ts a r eliabl e assa y t o de t ec t low PV AF s without the accumul a tion of f alse- positiv e signal. A t the inc r eas ed sensitivit y lev el, the assa y ’ s r eproducibility and repe at abili ty r at e w as >96% and pr ov ed t o be r o bus t t o op er ator , r ea g e n t lo t, and timepoi n t v aria bility . Comparison of t his PhasED-Seq-based MRD assa y ag ains t an orthogonal S NV-based appr oach f or de t ecti ng ctDNA-MRD demons trat ed a high ov er all c onc o r danc e. Th e disc or d ant c alls w e r e adjudic at e d ag ains t p a ti ent clinic al ou t co me dat a and f or all disc o r da n t cases, th e PhasED-Seq-based MRD assa y c orr el at ed with clini c al ou t c o me. Collectiv ely , the d at a prese n ted he r e sug g es ts that th e PhasED-Seq-based MRD assa y is accur a t e and r ep r oducible , making it appr op riate f or use in the clini c al setting f or individual s with B-cell malignancies. Thr ough t he dev el opment of impr ov ed ctDNA de t ecti on methods s uch as that pr es ent e d her e , patie n t ou t comes ma y be impr ov e d thr ough th e de t ecti on of r esidual dis ea se or early relapse which ma y be used t o guide t r e a tme n t d ecisions. All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprintthis version posted August 10, 2024. ; https://doi.org/10.1101/2024.08.09.24311742doi: medRxiv preprint 1 1 Tables and Figures Figure 1. Depiction of Informative Molecules. ‘Inf orma tiv e Molecul es’ are cfD NA molec ules spanning the lo c ation of a tumo r-specific PV . An y cfD NA molecules th a t could harb or a P V fr om the pa t ient ’ s P V lis t ar e c onsid er e d Inf ormative Mol ecules. ‘Mutant Molecul es’ are Inf ormative Mol e cules which harbor th e tumor-sp ecific mut a n t all ele of one o r mor e P V s in the patie n t ’ s P V Lis t. In this ex ample th e P V chr14:120T>C , chr14:130G>T spans 4 inf orma tive molecul es and 2 mut a n t mol ecules. 1 All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprintthis version posted August 10, 2024. ; https://doi.org/10.1101/2024.08.09.24311742doi: medRxiv preprint 12 T able 1. Overall B ack ground E rror R a t e Donor # replicates # PV Lists Interrogated Mutant Molecules Informative Molecules Backgroun d Error Rate (fraction) Total MRD Positive Calls Total Expected MRD Negative Calls False Positive Rate (%) 1 2 35 5 98,7 93 ,02 4 5.06E- 08 0 70 0 2 2 35 1 77,7 67 ,44 0 1.29E- 08 1 70 1.43 3 2 35 1 98,6 27 ,95 5 1.01E- 08 0 70 0 4 2 35 1 86,9 19 ,49 2 1.15E- 08 0 70 0 5 2 35 1 107, 26 6,7 08 9.32E- 09 0 70 0 6 2 35 2 100, 64 5,4 24 1.99E- 08 0 70 0 7 2 35 1 96,8 17 ,24 4 1.03E- 08 0 70 0 8 2 35 3 78,6 52 ,04 8 3.81E- 08 0 70 0 9 2 35 0 106, 24 3,1 84 0 0 70 0 10 2 35 2 117, 73 2,3 13 1.70E- 08 0 70 0 11 2 35 0 118, 47 4,9 36 0 0 70 0 12 2 35 1 103, 08 8,4 19 9.70E- 09 0 70 0 13 2 35 1 118, 45 0,1 45 8.44E- 09 0 70 0 14 2 35 1 95,0 51 ,10 6 1.05E- 08 0 70 0 15 2 35 5 114, 20 3,0 30 4.38E- 08 0 70 0 16 2 35 1 108, 91 8,2 82 9.18E- 09 0 70 0 17 2 35 2 92,9 32 ,85 8 2.15E- 08 0 70 0 18 2 35 2 101, 44 7,7 40 1.97E- 08 0 70 0 19 2 35 4 113, 34 5,5 78 3.53E- 08 0 70 0 20 2 35 1 100, 44 9,0 94 9.96E- 09 0 70 0 21 2 35 0 106, 47 3,1 94 0 0 70 0 22 2 35 1 111, 12 9,4 28 9.00E- 09 0 70 0 23 2 35 6 94,9 54 ,18 7 6.32E- 08 1 70 1.43 24 2 35 0 114, 46 1,5 52 0 0 70 0 25 2 35 4 95,1 07 ,44 4 4.21E- 08 0 70 0 26 2 35 3 106, 48 9,0 01 2.82E- 08 0 70 0 27 2 35 0 92,4 76 ,51 8 0 0 70 0 28 2 35 0 100, 07 4,7 91 0 0 70 0 29 2 35 0 101, 02 7,5 78 0 0 70 0 30 2 35 1 97,2 01 ,85 0 1.03E- 08 0 70 0 31 2 35 2 90,7 65 ,90 6 2.20E- 08 0 70 0 32 2 35 2 100, 14 8,1 87 2.00E- 08 1 70 1.43 33 2 35 4 96,1 88 ,44 0 4.16E- 08 0 70 0 34 2 35 0 97,4 27 ,93 9 0 0 70 0 35 2 35 6 103, 36 2,5 18 5.80E- 08 1 70 1.43 36 2 35 3 85,6 63 ,95 8 3.50E- 08 0 70 0 All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprintthis version posted August 10, 2024. ; https://doi.org/10.1101/2024.08.09.24311742doi: medRxiv preprint 13 Donor # replicates # PV Lists Interrogated Mutant Molecules Informative Molecules Backgroun d Error Rate (fraction) Total MRD Positive Calls Total Expected MRD Negative Calls False Positive Rate (%) 37 2 35 2 80,5 66 ,38 8 2.48E- 08 1 70 1.43 38 2 35 2 105, 45 0,4 56 1.90E- 08 0 70 0 39 2 35 1 91,2 65 ,49 5 1.10E- 08 0 70 0 40 2 35 4 82,9 89 ,86 8 4.82E- 08 0 70 0 41 2 35 2 76,5 68 ,24 4 2.61E- 08 0 70 0 42 2 35 1 103, 16 0,2 23 9.69E- 09 0 70 0 43 2 35 0 108, 66 8,9 63 0 0 70 0 44 2 35 2 90,1 03 ,24 3 2.22E- 08 0 70 0 45 2 35 3 95,6 41 ,81 4 3.14E- 08 0 70 0 46 2 35 2 101, 76 2,5 22 1.97E- 08 1 70 1.43 47 2 35 4 115, 43 4,6 70 3.47E- 08 0 70 0 48 2 35 3 101, 58 0,7 13 2.95E- 08 0 70 0 49 2 35 2 93,7 69 ,94 4 2.13E- 08 0 70 0 50 2 35 2 112, 61 3,4 43 1.78E- 08 1 70 1.43 51 2 35 0 86,9 35 ,98 1 0 0 70 0 52 2 35 4 102, 15 4,2 85 3.92E- 08 0 70 0 53 2 35 1 63,8 90 ,65 6 1.57E- 08 0 70 0 54 2 35 2 76,8 00 ,23 9 2.60E- 08 1 70 1.43 55 2 35 2 78,9 98 ,78 8 2.53E- 08 1 70 1.43 56 2 35 2 83,3 56 ,34 3 2.40E- 08 0 70 0 57 2 35 3 91,4 87 ,65 3 3.28E- 08 0 70 0 58 2 35 0 90,2 76 ,66 3 0 0 70 0 59 2 35 2 111, 06 5,7 06 1.80E- 08 0 70 0 60 2 35 2 121, 23 2,6 06 1.65E- 08 1 70 1.43 Overall 120 35 115 5,89 4, 553 ,41 7 1.95 E-0 8 10 4200 0.24 All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprintthis version posted August 10, 2024. ; https://doi.org/10.1101/2024.08.09.24311742doi: medRxiv preprint 14 T able 2. An aly tical Sensi tivit y Corrected PVAF Level n N Detection Rate (%) 4.83E- 06 10 10 100 2.42E- 06 10 10 100 1.21E- 06 10 10 100 6.04E- 07 9 10 90 3.02E- 07 5 10 50 1.51E- 07 2 10 20 0 0 4 0 LoD95 PVAF - Probit 6.61E- 07 PV AF , phased v ariant a llele fr act ion ; N , tot al r ep lic at es at le vel; n, MRD p ositive r esults; Mu ltip le l ots wer e used to ev aluat e the det ection r at es at each t ar get PV AF level. All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprintthis version posted August 10, 2024. ; https://doi.org/10.1101/2024.08.09.24311742doi: medRxiv preprint 15 T able 3. DLBCL MRD Assay Precision R esults Input Mass (ng) Numerator (f rom APA equation) Denominator (f rom APA equation) APA (%) 95% Lower and Upper CI Reprod ucib il ity 5 419 432.5 96.88 94.77, 98.3 Repeatability 60 62 96.77 79.19, 99.2 3 Reprod ucib il ity 120 149 149 100 97.55, 10 0 Repeatability 60 62 96.77 79.19, 99.2 3 ng, nanograms; CI, confidence interval; APA, average positive agreement; Analyses were performed utilizing different lots, operators, and timepoints to generate maximum variability. All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprintthis version posted August 10, 2024. ; https://doi.org/10.1101/2024.08.09.24311742doi: medRxiv preprint 16 T able 4. A c c urac y Betwee n-T es t Con c or da nce Comparator MRD Result NEGATIVE (n) POSITIVE (n) TOTAL (N) PhasED-Seq- based MRD

NEGATIVE (n) 14 3 17 POSITIVE (n) 4 29 33 TOTAL (N) 18 32 50 NPA 77.78% (95% CI 52.73%, 93.59%) PPA 90.62% (95% CI 74.98%, 98.02%) OPA 86.00% (95% CI 73.26%, 94.18%) All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprintthis version posted August 10, 2024. ; https://doi.org/10.1101/2024.08.09.24311742doi: medRxiv preprint 17 Table 5. Discordant Calls Donor Monitoring Timepoint PhasED-Seq- based MRD

Comparator MRD

Clinical Outcome MRD Call Discordance Explanation P1001 Baseline NEGATIVE POSITIVE Disease Free Patient cured ; MRD assay result matches clin ical o utcome. MRD assay result at EOT was MRD NEGATIVE. P1001 Cycle2Day1 NEGATIVE POSITIVE Disease Free Patient cured ; MRD assay result matches clin ical o utcome. MRD assay result at EOT was MRD NEGATIVE. P1002 Cycle2Day1 POSITIVE NEGATIVE Disease Free Bloo d co llect ion was dur ing treatment. Disease w as still detected at Cycle2Day1, based on MRD assay result. Patient had add iti onal cycles of the rapy that l ikely cleare d thei r disease. MRD as say result at Cycle3Day1 was MR D NEGATIVE wh ich matches c li nica l outcome. P1003 Cycle2Day1 POSITIVE NEGATIVE Disease Free Bloo d co llect ion was dur ing treatment. Disease w as still detected at Cycle2Day1, based on MRD assay result. Patient had add iti onal cycles of the rapy that l ikely cleare d thei r disease. MRD as say result at EOT was MRD NEGA TIVE match ing the cl in ical outcome. P1004 End O f Therapy NEGATIVE POSITIVE Disease Free Patient cured w ith >5 years fol low u p; MRD assay result at EO T matches cl ini cal outcome. P1005 Cycle3Day1 POSITIVE NEGATIVE Disease Free Bloo d co llect ion was dur ing treatment. Disease w as still detected at Cycle3Day1, based on MRD assay result. Patient had add iti onal cycles of the rapy that l ikely cleare d thei r disease. MRD as say result at EOT was MRD NEGA TIVE wh ich matches c lin ical outc ome. P1006 Cycle3Day1 POSITIVE NEGATIVE Disease Relapse Patient still has disease based on MRD assay resul t. MRD assa y result at EOT was MRD POSITIVE which matches cl inica l All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprintthis version posted August 10, 2024. ; https://doi.org/10.1101/2024.08.09.24311742doi: medRxiv preprint 18

(40 max) 1. Pessoa LS, Heringer M, Ferrer VP . Ctdna as a cancer biomark er: A broad overview . Crit Rev Oncol Hematol 2020;155:103109. Epub 20200925 doi: 10.1016/j.critrevonc.2020.103109. 2. Shields MD , Chen K, Dutcher G, Pate l I, Pellini B. Making the rounds: Exploring the role of circulating tumor DNA (ctdna) in non-small cell lung cancer . Int J Mol Sci 2022;23:16. Epub 20220812 doi: 10.3390/ijms23169006. 3. Anagnostou V , Ho C, Nicholas G, Juergens RA, Sacher A, Fung AS, et al. Ctdna response after pembrolizumab in non-small cell lung cancer: Phase 2 adaptive trial results. Nat Med 2023;29:10:2559-69. Epub 20231009 doi: 10.1038/s41591-023-02598-9. 4. Kim H, Park KU. Clinical circulating tumor DNA testing for precision oncology . Cancer Res T reat 2023;55:2:351-66 doi: 10.4143/crt.2022.1026. 5. Cronin KA, Scott S, Firth AU, Sung H, Henley SJ, Sherman RL, et al. Annual report to the nation on the status of cancer , part 1: National cancer statistics. Cancer 2022;128:24:4251-84. Epub 20221027 doi: 10.1002/cncr .34479. 6. Goldfinger M, Cooper DL. Refractory dlbcl: Challenges and treatment. Clin L ymphoma Myeloma Leuk 2022;22:3:140-8. Epub 20210920 doi: 10.1016/j.clml.2021.09.011. 7. Kostak oglu L, Martelli M, Sehn LH, Belada D , Carella A-M, Chua N, et al. End-of-treatment pet/ct predicts pfs and os in dlbcl after first-line treatment: Results from goya. Blood Advances 2021;5:5:1283-90 doi: 10.1182/bloodadvances.2020002690. 8. Lauer EM, Mutter J, Scherer F . Circulating tumor DNA in b-cell lymphoma: T echnical advances, clinical applications, and perspectives for translational research. Leuk emia 2022;36:9:2151-64 doi: 10.1038/s41375-022-01618-w . 9. Kurtz DM, Soo J, Co Ting Keh L, Alig S, Chabon JJ, Sworder BJ, et al. Enhanced detection of minimal residual disease by targeted sequencing of phased variants in circulating tumor DNA. Nat Biotechnol 2021;39:12:1537-47. Epub 20210722 doi: 10.1038/s41587-021- 00981-w . 10. E valuation of detection capability for clinical laboratory measurement procedures: Approved guideline. 2012. 11. Y u T , Li Q , Gray G, Y ue L Q. Statistical innovations in diagnostic device evaluation. Journal of Biopharmaceutical Statistics 2016;26:6:1067-77 doi: 10.1080/10543406.2016.1226332. 12. Scherer F , Kurtz DM, Newman AM, Stehr H, Craig AF , Esfahani MS, et al. Distinct biological subtypes and patterns of genome evolution in lymphoma revealed by circulating tumor DNA. Sci T ransl Med 2016;8:364:364r a155 doi: 10.1126/scitranslmed.aai8545. All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprintthis version posted August 10, 2024. ; https://doi.org/10.1101/2024.08.09.24311742doi: medRxiv preprint