Abstract Background: Programmed death-ligand 1 (PD-L1) expression, measured by tumor proportion score (TPS), guides immunotherapy (IO) selection in NSCLC. However, tissue-based PD-L1 immunohistochemistry (IHC) is often limited by insufficient tissue, sampling bias and intratumoral heterogeneity. cfDNA methylation signatures enable non-invasive measurement of tumor-derived epigenetic signals and may be able to capture PD-L1-associated biology from a simple blood draw. We developed a cfDNA methylation-based predictor to identify patients with low PD-L1 TPS (<50%), a group more likely to benefit from IO combination regimens rather than IO monotherapy. Methods: cfDNA methylation profiles of >500 plasma clinical patient samples, each with paired tumor PD-L1 IHC data, were analyzed across thousands of regulatory regions. A regularized logistic regression model was trained to predict samples with PD-L1 TPS <50%. Model performance was evaluated on an independent test cohort (N=90) by comparing predicted calls with IHC-based PD-L1 TPS measurements. Results: The cfDNA methylation predictor for identifying PD-L1 TPS <50% cases achieved >50% sensitivity, 87% specificity, and >90% positive prediction value (PPV). Model performance was consistent across NSCLC histologies (LUAD and LUSC) and remained robust at tumor fractions as low as 0.05%. Approximately 70% of liquid biopsy samples were evaluable, supporting the feasibility of cfDNA methylation analysis for the majority of clinical samples. Conclusions: Our methylation-based PD-L1 low predictor enables non-invasive detection of NSCLC cases with tissue PD-L1 TPS <50%, offering a potential alternative when tissue is limited or not available. Further investigation is warranted to determine whether an epigenetic PDL1 IHC trained classifier can support treatment decisions by identifying NSCLC patients more likely to require IO chemotherapy combination therapy. Citation Format: Wei Tian, Anton Valouev, Kunwar Singh, Matthew Ellis, Katie Quinn, Tingting Jiang, Martina Lefterova, Justin Odegaard, Darya Chudova. Non-invasive cfDNA methylation profiling for prediction of PD-L1 tumor proportion score status in NSCLC [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2026; Part 1 (Regular Abstracts); 2026 Apr 17-22; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2026;86(7 Suppl):Abstract nr 3842.

3070 Background: Short read lengths in next-generation sequencing and targeted panel probe design pose challenges to fusion detection by impairing the resolution of complex genomic rearrangements and the accurate mapping of intronic breakpoints. To address these limitations, we developed a cell-free DNA (cfDNA) methylation-based fusion epigenotyping method that leverages fusion-associated tumor epigenetic signatures to complement genomic-based methods and increase the fusion detection sensitivity of our liquid biopsy. We validated EML4 - ALK fusion detection by this algorithm in non-small cell lung cancer (NSCLC) using paired clinical tumor tissue and cfDNA samples. Methods: We trained a binary classifier to discriminate EML4-ALK fusion-positive from fusion-negative NSCLC using cfDNA methylation signal and genomic molecule support. To validate the epigenotyping classifier, an independent cohort of 577 clinical NSCLC samples was selected with paired tumor tissue and Guardant360 Liquid (Guardant Health, Palo Alto, CA) cfDNA for each patient (with epigenomic tumor fraction > 0.03%). The cohort included 94 tissue-confirmed fusion-positive samples (78 cfDNA genomic positives and 16 cfDNA genomic false negatives) with EML4-ALK detected in tumor tissue, and 483 tissue-confirmed fusion-negative samples ( ALK fusion negative in both tissue and paired cfDNA). The epigenotyping classifier predictions in cfDNA were evaluated against tissue-based orthogonal truth. Results: Among tissue-confirmed fusion-positive samples, tissue-liquid assessment showed a high positive percent agreement (PPA) / sensitivity of 89.36% (84/94), as well as 100% concordance with genomic caller positives (78/78). The rate of rescued fusions by the epigenotyping classifier from genomic false negative liquid cases was 38% (6/16). Among tissue-confirmed fusion-negative samples, tissue-liquid assessment showed a high negative percent agreement (NPA) / specificity of 99.38% (480/483). The false positive rate (FPR) of high confidence calls (probability exceeding a 99.7% specificity threshold or with partial genomic evidence) was 0.0% (0/483). Conclusions: cfDNA methylation-based fusion epigenotyping substantially increased detection of actionable ALK fusions while maintaining high specificity, as demonstrated by tissue-liquid concordance. The results of this approach showed the clinical value of giving NSCLC patients an increased likelihood to receive more effective, less toxic ALK inhibitor therapy, while the minimized FPR helps ensure appropriately matched treatment decisions.

Abstract Background: Fusion detection with next-generation sequencing is challenging due to short read fragments, which can fail to fully resolve complex genomic rearrangements and map intronic breakpoints that lie outside targeted capture regions. Tumor methylation patterns, which reflect the functional state of cancer cells and do not rely on breakpoint coverage, are less impacted by sequencing fragment length and provide a robust orthogonal signal to augment genomic-based fusion calling. We developed a cell-free DNA (cfDNA) methylation-based fusion epigenotyping method to rescue fusions missed by genomic-based methods, focusing on ALK fusion detection in non-small lung cancer (NSCLC) to inform ALK inhibitor therapy selection. Methods: NSCLC samples were processed using Guardant360 Liquid test (Guardant Health, Palo Alto, CA). Genome-wide cfDNA methylation profiles across thousands of regulatory regions together with genomic molecule support, were used to train a binary classifier that discriminates EML4-ALK fusion-positive from fusion-negative NSCLC. A logistic regression model was trained on 175 EML4-ALK fusion-positive and 175 fusion-negative samples. Concordance with NSCLC tissue samples was assessed by comparing tissue EML4-ALK-associated differentially methylated regions (DMRs, p<0.05) from TCGA data to fusion DMRs from Guardant360 Liquid cfDNA samples. To ensure clinical-grade specificity, a decision threshold targeting >99% specificity was calibrated on >11,000 genomic fusion-negative NSCLC samples. Model performance was evaluated on 102 independent positive cases (63 genomically detected, 39 genomically missed) with epigenomic tumor fraction > 0.1% from samples with (a) ALK inhibitor resistance mutations, (b) prior ALK inhibitor treatment, or (c) longitudinal history of genomic fusion detection. Fusion rescue rate was defined as the fraction of genomically missed fusions rescued by the epigenotyping classifier. Results: Methylation concordance analysis between TCGA NSCLC tissues and Guardant360 Liquid cfDNA samples showed significant overlap, with 64% (1384/2168) of tissue EML4-ALK-associated differentially methylated regions (DMRs, p<0.05) also significant in cfDNA. In the test cohort, the epigenotyping classifier achieved 74% sensitivity, detecting 100% of fusions identified by the genomic caller. Among genomically missed cases, the classifier rescued 31% (12/39) of fusions. Conclusion: A cfDNA methylation-based fusion epigenotyping approach provides a high-specificity orthogonal signal that augments genomic fusion detection, recovering a substantial fraction of EML4-ALK fusions missed by the genomic method. Clinically, patients with rescued ALK fusions could be considered for effective and low toxicity targeted ALK inhibitor therapies. Citation Format: Laura Tung, Anton Valouev, Justin Odegaard, Lauren Lawrence, Nicole Zhang, Martina Lefterova, Matthew Ellis, Tingting Jiang, Sheila Solomon, Darya Chudova, . Fusion epigenotyping using cell-free DNA methylation improves detection of actionable ALK fusions in non-small cell lung cancer [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2026; Part 1 (Regular Abstracts); 2026 Apr 17-22; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2026;86(7 Suppl):Abstract nr 1409.

Neoplastic progression in lung cancer often involves tumors with heterogeneous histologies, harboring multiple subclones of distinct histological subtypes - such as lung adenocarcinoma (LUAD), lung squamous cell carcinoma (LSCC), and small cell lung carcinoma (SCLC) - within a single tumor. Determining histological subtype of the tumor typically requires invasive lung tissue biopsy, which precludes serial monitoring both during and after therapy. Treatment regimens targeting the predominant histology may inadvertently select for drug-resistant subclones of alternative subtypes, necessitating timely adjustments in treatment strategies. Comprehensive cfDNA methylation profiling across thousands of cancer-associated regulatory regions offers an approach for accurate quantification of lung cancer histological subtypes, facilitating enhanced noninvasive disease monitoring and more effective therapeutic strategies. We developed a novel cfDNA methylation-based deconvolution model to identify and quantify the proportional contributions of LUAD, LSCC and SCLC in patient blood samples. The model leverages hypermethylation signals across over 9000 of cancer-associated genomic regions to estimate the contribution of each histological subtype while simultaneously predicting patient-specific subtype hypermethylation signatures and non-cancer component signatures across regions. We modeled each sample as a linear combination of its three histological subtype and non-cancer components weighted by their proportional contributions. The model was trained using stochastic gradient descent to minimize both reconstruction and proportional contribution losses. Using samples from cancer-free individuals (N=6, 142), and lung cancer patients (LUAD N=2, 721, LSCC N=663, and SCLC N=243), the model was trained using 90% of the data and tested on the remaining 10%. For each test sample, the model was used to predict proportions of three histological subtypes summing to 100%. Model performance was evaluated by comparing the predicted histological subtypes to the known histological subtype annotations of lung cancer samples. The overall accuracy on the test set was 85.1% down to 0.1% tumor fraction, with the highest accuracy for LUAD (90.4%), followed by LSCC (72.4%) and SCLC (63.5%). Importantly, the model demonstrated capability to detect minor subclonal contributions as low as 0.1%. Our methylation-based subtype deconvolution model provides an accurate method for quantifying proportions of lung cancer histological subtypes in patient blood samples, with the unique ability to detect minor subclonal populations. This approach enables the early non-invasive detection of treatment-resistant lung cancer subtypes, thereby holding significant potential to improve patient outcomes. Anton Valouev, Wei Tian, Kunwar Singh, Sheila Solomon, Justin Odegaard, Darya Chudova, Helmy Eltoukhy. Non-invasive cell free DNA (cfDNA) methylation profiling for accurate proportional quantification of lung cancer subtypes [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2025; Part 1 (Regular Abstracts); 2025 Apr 25-30; Chicago, IL. Philadelphia (PA): AACR; Cancer Res 2025;85(8_Suppl_1):Abstract nr 1142.

68 Background: Blood based colorectal cancer (CRC) screening has been validated for use in average risk populations (ECLIPSE NCT#04136002; Guardant Health, USA). Here we present the performance of an enhanced version of a blood-based screening test developed to optimize detection of low shedding tumors by leveraging epigenomic features of cell-free DNA (cfDNA). Test performance was assessed across screen-relevant individuals from the intended use population, including a collection of screen-detected CRC and enriched cohort of individuals with endoscopy finding of CRC. Methods: We trained a regression model to classify whether aberrant cfDNA originated from individuals with CRC or non-ACN individuals and compared performance to an original model in a fixed set of screen-relevant samples sequenced to a median of 11M reads per sample across a panel approximately 1Mb in size. Model optimization focused on identifying training settings and samples to maximize detection of low shedding screen-relevant tumors and introduced noise reduction techniques to minimize technical variation. The final enhanced model was trained in over 4500 samples; the calling cut-off was set targeting 90% specificity in the average-risk population following US-Census age distribution. Analytical Limit of Detection (LoD) was estimated using ~5,000 in silico dilution samples generated from blood samples from 25 CRCs and ~1,650 non-ACNs. Results: This training approach yielded a 2X improvement in the analytical LoD95 (0.004% vs 0.008%) compared to the original model validated in the ECLIPSE trial. The model also yielded an increase in overall CRC clinical sensitivity from 84% to 91%, N=45, with notable sensitivity improvement from 76% to 88% in detecting early localized stage I/II CRC, N = 25, while maintaining specificity at 91%. Conclusions: This enhanced cfDNA-based CRC screening blood test shows improved performance in early stage CRC detection demonstrating the potential of continuous improvement in the performance of cfDNA-based screening tests powered by data and clinical insights.

Abstract Background. Multi-cancer blood-based tests may yield clinical benefit by improving compliance to guideline recommended screening with a more patient-friendly modality, and also by detection of early (stage I/II) tumors in cancer types that lack screening tests, yet early intervention can save lives. A single test with clinically meaningful performance which addresses both opportunities has yet to be developed. Methods. We evaluated a blood-based multi-modal device based on cfDNA epigenomic and targeted protein analysis that enables high performance for early-stage cancer detection in colon, lung, bladder, gastric, liver, ovarian, and pancreas cancers. Blood samples were obtained from multiple case-control cohorts of individuals with colorectal (N > 2,000), lung (N > 300), and other solid tumor cancers (bladder, gastric, liver, ovarian, pancreas (N > 300)) as well as individuals without cancer (N > 3,000). The assay is based on Guardant Shield™ blood-based CRC screening test and the bioinformatic pipeline is augmented with a multi-cancer screening caller for the detection of additional cancers. Sensitivity for CRC and lung cancer detection is calculated at 90% target specificity thresholds, due to availability of guideline recommended screening tests, colonoscopy and LDCT scan to adjudicate blood-based test results. Specificity for all other cancers were targeted at an overall specificity of 98%. The specificity thresholds for CRC, lung, and multi-cancer are selected to yield assay performance tailored for the cancer type and clinical diagnostic pathway. Results. In this study cohort, this integrated, single device, multi-cancer test yielded CRC sensitivity of 91% (stage I/II: 93%) and lung cancer sensitivity of 85% (stage I/II: 75%) at 90% specificity. The overall sensitivity in bladder, gastric, liver, ovarian, and pancreas cancers was 75% (stage I/II: 66%) at 98% overall specificity. Overall prevalence-adjusted sensitivity of this device in the above 7 cancers were 79% overall sensitivity (stage I/II: 78%). This blood-based test could detect 25% of the expected cancer diagnoses in 2022 according to SEER estimates. Conclusions. This highly-sensitive, integrated, blood-based cancer screening device yields performance on par with currently available screening tests for cancers with screening guidelines (CRC and lung) and clinically meaningful early-stage detection in cancer types without screening guidelines where early intervention can bring clinical benefit, highlighting the ability of this technology to yield clinically meaningful results for the detection of early stage cancer. The performance of this device is under investigation in prospective screening trials. Citation Format: Yupeng He, Anton Valouev, Liyang Xiong, William W. Young Greenwald, Victoria M. Raymond, Sven Duenwald, AmirAli Talasaz. Highly sensitive blood-based multi-cancer screening device with tiered specificity based on diagnostic workflow [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2023; Part 1 (Regular and Invited Abstracts); 2023 Apr 14-19; Orlando, FL. Philadelphia (PA): AACR; Cancer Res 2023;83(7_Suppl):Abstract nr 3331.

Background: Liquid biopsy offers a rapid and non-invasive alternative to tissue biopsy for identifying biomarkers. More recently, its application has broadened to include assessment of early response to therapy (i.e. molecular response) and in the early-stage settings, detection of minimal residual disease (MRD) and early disease recurrence1. While circulating tumor fraction (cTF) estimated by somatic mutations is well associated with the tumor progression and prognosis, interference can occur from clonal hematopoiesis of indeterminate potential (CHIP), and for cell-free DNA (cfDNA) samples that lack detectable somatic mutations, somatic tumor fraction cannot be estimated. In this analysis, we demonstrate that epigenomic signatures accurately measure cTF using orthogonal analytes to somatic mutations and enable cTF estimation even in cases without detectable tumor driver variants. Methods: To capture tumor-associated methylated cfDNA, we designed a custom assay of a broad genomic panel (15.2 Mb) targeting unmethylated regions in plasma cfDNA from healthy individuals. We profiled plasma samples from cancer patients with this panel, and utilized machine learning to integrate methylation signals into an estimate of cTF. We benchmarked the accuracy of methylation cTFs on real plasma samples, as well as in-vitro and in-silico titration datasets. Both titration data sets were generated by mixing cfDNA from patients with colorectal cancer (CRC) into the plasma from cancer-free donors, either via titration of CRC cfDNA into cfDNA from cancer-free donors for the in-vitro data, or via computationally mixing reads from CRC patients with those from cancer-free donors for the in-silico data. Results: Our methylation cTF quantified a similar cTF to those derived from well-calibrated genomic tumor driver mutations; across the 670 stage I-IV CRC samples, a strong correlation (Pearson r=0.85) was observed between methylation logit(cTF) and genomic logit(cTF). The methylation cTF was capable of quantifying low cTFs: it quantified a cTF over 0.1% in >99% of the 270 in-vitro and 1,000 in-silico titration samples with true cTFs >0.1%. In contrast, when applied to 2,037 cancer-free samples, less than 5% of the samples resulted in estimated cTFs of >0.1%. Our methylation cTF was more robust than genomic cTF on the 62 in vitro titration samples with true cTFs between 0.3-1%, with a five fold lower coefficient of variation across methylation cTFs compared to genomic cTFs. Conclusions: cTFs from methylated cfDNA may overcome the current limitations of somatic mutation based methods. Our methylation approach is capable of accurately detecting cTFs in tumor-driver positive and negative cases. As we estimate tumor-negative cases to be 30-50% of patients with stage I-III cancer and 15-20% of patients with stage IV cancer, our methylation approach may hold promise for providing better evaluation for patient care and management. Citation Format: William W. Greenwald, Yupeng He, Sai Chen, Tingting Jiang, Anton Valouev, Jun Min, Catalin Barbacioru, Daniel P. Gaile, Dustin Ma, Yvonne Kim, Giao Tran, Indira Wu, Ariel Jaimovich, Victoria Raymond, Rebecca J. Nagy, Han-Yu Chuang. Accurate epigenomic estimates of circulating tumor fraction in large-scale clinical data [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2022; 2022 Apr 8-13. Philadelphia (PA): AACR; Cancer Res 2022;82(12_Suppl):Abstract nr 3758.

Background: A blood-based cancer screening test must exhibit performance metrics optimized for the cancer of interest based on associated clinical diagnostic pathways and demonstrate an ability to detect disease at an early stage when intervention has a meaningful impact on individual and net population health outcomes. We evaluated the performance of a blood-based cancer screening assay in select tumor types where we believe cancer screening can save lives. The assay interrogates cell-free DNA (cfDNA) methylation signatures for early-stage cancer (stage I/II) detection and tissue of origin identification. Methods: Whole blood samples from individuals with (N > 1,500) and without (N > 1,800) cancer were obtained from multiple cohorts. Plasma-derived cfDNA was profiled using a custom assay that enriches fragments with dense CpG methylation and further depletes uninformative background molecules. A broad genomic panel (16 Mb) targeting regions with low rates of methylation in healthy individuals was used to capture and sequence tumor-associated molecules while maintaining high sensitivity at low sequencing cost per sample. A cross-validated analysis was used to estimate out of sample performance of the predictive model. Classification thresholds corresponding to 90% and 95% specificities were established using a set of samples from individuals without a cancer diagnosis. Results: To evaluate the performance of this screening assay in cancers with guideline-directed screening protocols, colorectal and lung cancers, detection was assessed at 90% specificity. At this threshold, sensitivity for stage I/II colorectal and lung cancer was 90% and 87%, respectively. For other cancers with no current guideline-directed screening paradigms, pancreatic and bladder cancers, a specificity threshold of 95% was applied. Sensitivity was 73% and 52% for stage I/II pancreatic and bladder cancer, respectively. Tissue of origin prediction evaluated at 98% specificity had accurate identification in 99% of colorectal, 94% of lung, 88% of bladder, and 86% of pancreatic cancers. Conclusions: This multi-cancer targeted screening assay provides robust and sensitive detection of early-stage cancer at thresholds optimized for current screening paradigms with accurate tissue of origin identification. The assay is undergoing further expansion of its detection capabilities to include additional cancer types where screening can save lives. Clinical evaluation in registrational screening trials is ongoing (NCT05117840). Citation Format: Anton Valouev, Elena Zotenko, Matthew Snyder, Charbel Eid, Ngan Nguyen, Jun Min, Yupeng He, Ariel Jaimovich, Haley Axelrod, Prashanthi Natarajan, Anna Hartwig, Noam Vardi, Tam Banh, Andrew Kennedy, William Greenleaf, Stefanie Mortimer, Sven Duenwald, Darya Chudova, AmirAli Talasaz. Development of a highly-sensitive targeted cell-free DNA epigenomic assay for early-stage multi-cancer screening [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2022; 2022 Apr 8-13. Philadelphia (PA): AACR; Cancer Res 2022;82(12_Suppl):Abstract nr 2141.

3542 Background: Cancer screening in asymptomatic individuals who meet guideline criteria has yielded reductions in cancer death rates. However, adherence to screening guidelines remains below targets set forth by leading health-care organizations. A blood-based multi-cancer screening assay with clinically meaningful sensitivity and specificity, in cancer types where early detection and intervention can save lives, that is integrated with existing clinical pathways may increase access to and adherence with guideline recommendations, ensuring more individuals benefit from these proven interventions. We evaluated the performance of a blood-based multi-cancer screening assay that interrogates cell-free DNA (cfDNA) methylation signatures for cancer detection and tissue of origin prediction in a set of tumor types where cancer screening can save lives. Methods: Whole blood from 1,607 individuals with and 3,298 individuals without cancer was obtained from multiple unique cohorts. Plasma-derived cfDNA was profiled using a custom assay that enriches fragments with dense CpG methylation and further depletes uninformative background molecules containing unmethylated CpGs. We utilized a broad genomic panel (16 Mb) targeting regions with low rates of methylation in individuals without cancer. The panel captures tumor-associated molecules and allows for high sensitivity of detection at low sequencing costs. A cross-validated analysis was used to estimate the performance of the predictive model upon the sample set. Classification thresholds corresponding to 90%, 95%, and 98% specificities were established using samples from individuals without a cancer diagnosis. Results: At 90% specificity, overall sensitivity for lung cancer detection was 92.1% (95% CI: 80-100%; 90.2% in Stage I/II disease (N = 82) and 93.1% in Stage III/IV disease (N = 159)) and 93.1% (CI: 88-98%) for CRC detection (92% in Stage I/II disease (N = 743) and 94.5% in Stage III/IV disease (N = 623)). Tissue of origin prediction evaluated at 98% specificity yielded accurate identification in 99% of CRC and 98% of lung cancers. Lung cancer histology was known for approximately 74% of the cohort. Across Stage I – IV cancers, at 90% specificity, sensitivity was 97.3% in lung squamous cancer (N = 73) and 86.8% in lung adenocarcinoma (N = 106). At 95% and 98% specificity thresholds, overall sensitivity was 86.3% (CI: 75-98%) and 66.4% (CI: 56-77%) for lung cancer and 85.7% (CI:81-91%) and 71.6% (CI: 67-76%) for CRC, respectively. Conclusions: This blood-based multi-cancer screening assay yields clinically meaningful sensitivity and specificity for early-stage cancers. This assay is undergoing further development to expand detection capabilities to additional cancer types where screening can save lives. Clinical evaluation in registrational screening trials is ongoing (SHIELD; NCT05117840).

536 Background: New breast cancer screening approaches are needed to detect clinically aggressive subtypes that may not be detected by mammography or are detected late in unscreened populations. CCGA (NCT02889978) is a prospective multi-center observational study for the development of a noninvasive assay for cancer detection. A preplanned substudy of a Women-Only Cohort is reported. Methods: Blood was prospectively collected (N = 1627) from 878 participants (pts) with newly diagnosed untreated cancer (20 tumor types, all stages) and 749 pts with no cancer diagnosis (controls, C) for plasma cfDNA extraction. This substudy included 358 pts with invasive breast cancer (IBC) and 452 C. Three prototype sequencing assays were performed: paired cfDNA and white blood cell (WBC) targeted sequencing (507 genes, 60,000X) for single nucleotide variants/indels, paired cfDNA and WBC whole genome sequencing (WGS, 30X) for copy number variation, and cfDNA whole genome bisulfite sequencing (WGBS, 30X) for methylation; WBC sequencing identified the contribution of clonal hematopoiesis (CH). For each assay, a classification model using 10-fold cross-validation was developed to discriminate IBC from C using a subset of women; sensitivity was estimated at 95% specificity. Results: IBC pts and C had similar age (mean yrs±SD: 58±13 IBC, 59±12 C). 46% of IBC pts were symptomatic (a subset were documented interval cancers), 82% were stage I/II. The subtype breakdown of HR+/HER2+/triple-negative breast cancer (TNBC) was 65%/17%/15%. WGBS returned the highest sensitivity of the 3 assays and is reported here; results were consistent across all assays. Sensitivity (95% CI) was higher for TNBC vs HER2+ vs HR+/HER2- (58% [43-72] vs 40% [28-54] vs 15% [10-20]), and higher for symptomatic vs screen-detected breast cancer (44% [36-52] vs 10% [6-16]). Comparison to tumor WGS and multi-assay classification will be reported. Conclusions: Breast cancers with detectable cfDNA signals at time of diagnosis included clinically aggressive subtypes and symptomatic presentation. Further assay and clinical development in the intended use population is ongoing (NCT03085888). Clinical trial information: NCT02889978.

Abstract CCGA [NCT02889978] is the largest study of cfDNA-based early cancer detection; the first CCGA learnings from multiple cfDNA assays are reported here. This prospective, multi-center, observational study has enrolled 10,012 of 15,000 demographically-balanced participants at 141 sites. Blood was collected from participants with newly diagnosed therapy-naive cancer (C, case) and participants without a diagnosis of cancer (noncancer [NC], control) as defined at enrollment. This preplanned substudy included 878 cases, 580 controls, and 169 assay controls (n=1627) across 20 tumor types and all clinical stages. All samples were analyzed by: 1) Paired cfDNA and white blood cell (WBC)-targeted sequencing (60,000X, 507 gene panel); a joint caller removed WBC-derived somatic variants and residual technical noise; 2) Paired cfDNA and WBC whole-genome sequencing (WGS; 35X); a novel machine learning algorithm generated cancer-related signal scores; joint analysis identified shared events; and 3) cfDNA whole-genome bisulfite sequencing (WGBS; 34X); normalized scores were generated using abnormally methylated fragments. In the targeted assay, non-tumor WBC-matched cfDNA somatic variants (SNVs/indels) accounted for 76% of all variants in NC and 65% in C. Consistent with somatic mosaicism (i.e., clonal hematopoiesis), WBC-matched variants increased with age; several were non-canonical loss-of-function mutations not previously reported. After WBC variant removal, canonical driver somatic variants were highly specific to C (e.g., in EGFR and PIK3CA, 0 NC had variants vs 11 and 30, respectively, of C). Similarly, of 8 NC with somatic copy number alterations (SCNAs) detected with WGS, 4 were derived from WBCs. WGBS data revealed informative hyper- and hypo-fragment level CpGs (1:2 ratio); a subset was used to calculate methylation scores. A consistent “cancer-like” signal was observed in <1% of NC participants across all assays (representing potential undiagnosed cancers). An increasing trend was observed in NC vs stages I-III vs stage IV (nonsyn. SNVs/indels per Mb [Mean±SD] NC: 1.01±0.86, stages I-III: 2.43±3.98; stage IV: 6.45±6.79; WGS score NC: 0.00±0.08, I-III: 0.27±0.98; IV: 1.95± 2.33; methylation score NC: 0±0.50; I-III: 1.02±1.77; IV: 3.94±1.70). These data demonstrate the feasibility of achieving >99% specificity for invasive cancer, and support the promise of cfDNA assay for early cancer detection. Additional data will be presented on detected plasma:tissue variant concordance and on multi-assay modeling. Citation Format: Alexander A. Aravanis, Geoffrey R. Oxnard, Tara Maddala, Earl Hubbell, Oliver Venn, Arash Jamshidi, Ling Shen, Hamed Amini, John A. Beausang, Craig Betts, Daniel Civello, Konstantin Davydov, Saniya Fazullina, Darya Filippova, Sante Gnerre, Samuel Gross, Chenlu Hou, Roger Jiang, Byoungsok Jung, Kathryn Kurtzman, Collin Melton, Shivani Nautiyal, Jonathan Newman, Joshua Newman, Cosmos Nicolaou, Richard Rava, Onur Sakarya, Ravi Vijaya Satya, Seyedmehdi Shojaee, Kristan Steffen, Anton Valouev, Hui Xu, Jeanne Yue, Nan Zhang, Jose Baselga, Rosanna Lapham, Daron G. Davis, David Smith, Donald Richards, Michael V. Seiden, Charles Swanton, Timothy J. Yeatman, Robert Tibshirani, Christina Curtis, Sylvia K. Plevritis, Richard Williams, Eric Klein, Anne-Renee Hartman, Minetta C. Liu. Development of plasma cell-free DNA (cfDNA) assays for early cancer detection: first insights from the Circulating Cell-Free Genome Atlas Study (CCGA) [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2018; 2018 Apr 14-18; Chicago, IL. Philadelphia (PA): AACR; Cancer Res 2018;78(13 Suppl):Abstract nr LB-343.

LBA8501 Background: Plasma cfDNA genomic analysis is used widely for the care of advanced lung cancer, but its suitability for early stage lung cancer detection is not well established. CCGA (NCT02889978) is a prospective, multi-center, observational study launched for the development of a noninvasive assay for cancer detection. Methods: Blood was prospectively collected (N = 1627) from 749 controls (no cancer diagnosis) and 878 participants (pts) with newly-diagnosed untreated cancer in this preplanned substudy, including 127 pts with lung cancer. Three prototype sequencing assays were performed: paired cfDNA and white blood cell (WBC) targeted sequencing (507 genes, 60,000X) for single nucleotide variants/indels; paired cfDNA and WBC whole genome sequencing (WGS) for copy number variation (30X); and cfDNA whole genome bisulfite sequencing (WGBS) for methylation (30X). For each assay, a classification model using 10-fold cross-validation was developed for all pts with cancer, then evaluated in the pts with lung cancer; sensitivity was estimated at 95% specificity. Results: We evaluated pts with lung cancer (127) and a subset of controls (580) with similar ages (mean±SD yrs: 67±9, 60±13), 85% and 43% were ever-smokers, and 46% and 22% were men, respectively. Of 3055 nonsynonymous mutations detected across 122 evaluable pts with lung cancer, > 50% were detected in WBC consistent with clonal hematopoiesis (CH). Accounting for CH, sensitivity in 63 stage I-IIIA pts evaluable across all 3 assays was 48% (35-61, targeted), 54% (41-67, WGS), and 56% (43-68, WGBS); in 54 stage IIIB-IV pts it was 85% (73-93, targeted), 91% (80-97, WGS), and 93% (82-98, WGBS) . Similar sensitivities were observed across histological subtypes (adenocarcinoma, squamous cell, small cell). Comparison to tumor WGS and multi-assay classification will be reported. Conclusions: Early stage lung cancers are detectable in cfDNA using a genome-wide sequencing approach. For lung cancer detection using targeted assays, CH must be accounted for to minimize false positives. Assay optimization is ongoing to allow further clinical development in the intended use population. Clinical trial information: NCT02889978.

12021 Background: Globally most cancers are detected at advanced stages with high treatment burden and low cure rates. A noninvasive cfDNA blood test detecting multiple cancers at early stages when curative treatment is more likely to succeed is desirable. CCGA (NCT02889978) is a prospective multi-center observational study for development of a noninvasive cfDNA-based multi-cancer detection assay. Methods: Prospectively collected samples (N = 1627) from 749 controls (no cancer diagnosis, C) and 878 participants (pts) with newly diagnosed untreated cancer (20 tumor types, all stages) were analyzed in a preplanned substudy. 3 prototype sequencing assays were performed: paired cfDNA and white blood cell (WBC, 60,000X) targeted sequencing (507 genes) for single nucleotide variants/indels; paired cfDNA and WBC whole genome sequencing (WGS, 30X) for copy number variation; cfDNA whole genome bisulfite sequencing (WGBS, 30X) for methylation. For each assay a detection model was developed for all cancer pts; sensitivity was estimated at 95% specificity. Results: Pts w/cancer and C had similar age, smoking status and gender. WGBS had the highest sensitivity and is reported here; results were consistent across assays. Detected (sensitivity [95% CI]) cancers (stage I-III) included 28 colorectal (66% [48-84]), 19 esophageal (63% [38-84]), 5 head and neck (56% [21-86]), 5 hepatobiliary (80% [28-99]), 73 lung (59% [47-70]), 17 lymphoma (77% [50-93]), 11 multiple myeloma (73% [39-94]), 10 ovarian (90% [56-99]), and 10 pancreatic (80% [44-98]). Breast cancer-specific assay results are reported separately. Cancers with low signal ( < 10% sensitivity) include low gleason score prostate cancer, thyroid, uterine, melanoma, and renal. Comparison to tumor WGS and multi-assay classification will be reported. Conclusions: A cfDNA-based blood test detected multiple cancers at various stages with high specificity, indicating this approach is promising as a multi-cancer screening test, including for lethal unscreened cancers where stage shift can impact mortality. Further assay and clinical development of a multi-cancer cfDNA test in an asymptomatic population is ongoing. Clinical trial information: NCT02889978.

12003 Background: CHIP is defined by the presence of age-dependent acquired mutations in hematopoietic progenitor cells and has been reported to occur in up to 30% of individuals 60-70 years of age. CHIP is a risk factor for hematologic malignancies and cardiovascular disease; its biological mechanisms and clinical significance are just now being studied. Using an assay ~100X more sensitive than exome sequencing, we determined the prevalence and features of CHIP in the CCGA cohort, and the impact on interpretation of cell-free DNA (cfDNA) somatic variants. Methods: Blood was prospectively collected (N = 1627) from 749 controls (no cancer, C) and 878 participants (pts) with newly-diagnosed untreated cancer (20 tumor types, all stages) for WBC and cfDNA isolation. Paired white blood cell (WBC) and cfDNA targeted sequencing (507 genes, 60,000X median coverage) identified somatic single nucleotide variants/indels. Unique molecular barcodes and a machine learning-based noise model achieved a specificity of 1 false positive variant call per Mb of genome targeted at a limit of detection of ~0.1% variant allele frequency (VAF). Results: 1412 samples were eligible and evaluable (576 C, 836 pts; 18 solid tumor types, all stages). Of somatic cfDNA variants matched in WBC (CHIP), 7% of individuals had CHIP with VAF > 10%, 39% had CHIP with VAF > 1%, and nearly all pts (92%) had a somatic mutation with VAF > 0.1%. The rate was similar between C and pts (median age 62, 60), increasing in prevalence by 160% per decade, such that we observed 2.5 variants/Mb at age 60. Of CHIP variants identified, 92% were unique to individual patients, most of which were present at low VAF. Genes impacted by CHIP included DNMT3A (40%), TET2 (27%), and TP53 (10%), consistent with previous reports in patients with solid tumors. Conclusions: An ultra-sensitive sequencing assay demonstrated that CHIP signal in WBC and cfDNA is much more common than previously appreciated. The clinical significance of CHIP warrants further study and must be accounted for when interpreting cfDNA variants for both early cancer detection and tumor genotyping (liquid biopsy). Clinical trial information: NCT02889978.

Nephron progenitor number determines nephron endowment; a reduced nephron count is linked to the onset of kidney disease. Several transcriptional regulators including Six2, Wt1, Osr1, Sall1, Eya1, Pax2, and Hox11 paralogues are required for specification and/or maintenance of nephron progenitors. However, little is known about the regulatory intersection of these players. Here, we have mapped nephron progenitor-specific transcriptional networks of Six2, Hoxd11, Osr1, and Wt1. We identified 373 multi-factor associated 'regulatory hotspots' around genes closely associated with progenitor programs. To examine their functional significance, we deleted 'hotspot' enhancer elements for Six2 and Wnt4. Removal of the distal enhancer for Six2 leads to a ~40% reduction in Six2 expression. When combined with a Six2 null allele, progeny display a premature depletion of nephron progenitors. Loss of the Wnt4 enhancer led to a significant reduction of Wnt4 expression in renal vesicles and a mildly hypoplastic kidney, a phenotype also enhanced in combination with a Wnt4 null mutation. To explore the regulatory landscape that supports proper target gene expression, we performed CTCF ChIP-seq to identify insulator-boundary regions. One such putative boundary lies between the Six2 and Six3 loci. Evidence for the functional significance of this boundary was obtained by deep sequencing of the radiation-induced Brachyrrhine (Br) mutant allele. We identified an inversion of the Six2/Six3 locus around the CTCF-bound boundary, removing Six2 from its distal enhancer regulation, but placed next to Six3 enhancer elements which support ectopic Six2 expression in the lens where Six3 is normally expressed. Six3 is now predicted to fall under control of the Six2 distal enhancer. Consistent with this view, we observed ectopic Six3 in nephron progenitors. 4C-seq supports the model for Six2 distal enhancer interactions in wild-type and Br/+ mouse kidneys. Together, these data expand our view of the regulatory genome and regulatory landscape underpinning mammalian nephrogenesis.

Siberia and Western Russia are home to over 40 culturally and linguistically diverse indigenous ethnic groups. Yet, genetic variation of peoples from this region is largely uncharacterized. We present whole-genome sequencing data from 28 individuals belonging to 14 distinct indigenous populations from that region. We combine these datasets with additional 32 modern-day and 15 ancient human genomes to build and compare autosomal, Y-DNA and mtDNA trees. Our results provide new links between modern and ancient inhabitants of Eurasia. Siberians share 38% of ancestry with descendants of the 45,000-year-old Ust-Ishim people, who were previously believed to have no modern-day descendants. Western Siberians trace 57% of their ancestry to the Ancient North Eurasians, represented by the 24,000-year-old Siberian Malta boy. In addition, Siberians admixtures are present in lineages represented by Eastern European hunter-gatherers from Samara, Karelia, Hungary and Sweden (from 8,000-6,600 years ago), as well as Yamnaya culture people (5,300-4,700 years ago) and modern-day northeastern Europeans. These results provide new evidence of ancient gene flow from Siberia into Europe.

Haifan Lin, Jamy Peng and colleagues report that the Drosophila Piwi protein is a negative regulator of PRC2 in the fly ovary, a function required for the maintenance of germline stem cells. Their results indicate that Piwi sequesters PRC2 in the nucleoplasm, thereby reducing genome-wide levels of H3K27me3. The Drosophila melanogaster Piwi protein regulates both niche and intrinsic mechanisms to maintain germline stem cells, but its underlying mechanism remains unclear. Here we report that Piwi interacts with Polycomb group complexes PRC1 and PRC2 in niche and germline cells to regulate ovarian germline stem cells and oogenesis. Piwi physically interacts with the PRC2 subunits Su(z)12 and Esc in the ovary and in vitro. Chromatin coimmunoprecipitation of Piwi, the PRC2 enzymatic subunit E(z), histone H3 trimethylated at lysine 27 (H3K27me3) and RNA polymerase II in wild-type and piwi mutant ovaries demonstrates that Piwi binds a conserved DNA motif at ∼72 genomic sites and inhibits PRC2 binding to many non-Piwi-binding genomic targets and H3K27 trimethylation. Moreover, Piwi influences RNA polymerase II activities in Drosophila ovaries, likely via inhibiting PRC2. We hypothesize that Piwi negatively regulates PRC2 binding by sequestering PRC2 in the nucleoplasm, thus reducing PRC2 binding to many targets and influencing transcription during oogenesis.

Nephron endowment is determined by the self-renewal and induction of a nephron progenitor pool established at the onset of kidney development. In the mouse, the related transcriptional regulators Six1 and Six2 play non-overlapping roles in nephron progenitors. Transient Six1 activity prefigures, and is essential for, active nephrogenesis. By contrast, Six2 maintains later progenitor self-renewal from the onset of nephrogenesis. We compared the regulatory actions of Six2 in mouse and human nephron progenitors by chromatin immunoprecipitation followed by DNA sequencing (ChIP-seq). Surprisingly, SIX1 was identified as a SIX2 target unique to the human nephron progenitors. Furthermore, RNA-seq and immunostaining revealed overlapping SIX1 and SIX2 activity in 16 week human fetal nephron progenitors. Comparative bioinformatic analysis of human SIX1 and SIX2 ChIP-seq showed each factor targeted a similar set of cis-regulatory modules binding an identical target recognition motif. In contrast to the mouse where Six2 binds its own enhancers but does not interact with DNA around Six1, both human SIX1 and SIX2 bind homologous SIX2 enhancers and putative enhancers positioned around SIX1. Transgenic analysis of a putative human SIX1 enhancer in the mouse revealed a transient, mouse-like, pre-nephrogenic, Six1 regulatory pattern. Together, these data demonstrate a divergence in SIX-factor regulation between mouse and human nephron progenitors. In the human, an auto/cross-regulatory loop drives continued SIX1 and SIX2 expression during active nephrogenesis. By contrast, the mouse establishes only an auto-regulatory Six2 loop. These data suggest differential SIX-factor regulation might have contributed to species differences in nephron progenitor programs such as the duration of nephrogenesis and the final nephron count.

Drosophila Piwi was reported by Huang et al. (2013) to be guided by piRNAs to piRNA-complementary sites in the genome, which then recruits heterochromatin protein 1a and histone methyltransferase Su(Var)3-9 to the sites. Among additional findings, Huang et al. (2013) also reported Piwi binding sites in the genome and the reduction of RNA polymerase II in euchromatin but its increase in pericentric regions in piwi mutants. Marinov et al. (2015) disputed the validity of the Huang et al. bioinformatic pipeline that led to the last two claims. Here we report our independent reanalysis of the data using current bioinformatic methods. Our reanalysis agrees with Marinov et al. (2015) that Piwi's genomic targets still remain to be identified but confirms the Huang et al. claim that Piwi influences RNA polymerase II distribution in the genome. This Matters Arising Response addresses the Marinov et al. (2015) Matters Arising, published concurrently in this issue of Developmental Cell.

Discovery of lineage-specific somatic copy number variation (CNV) in mammals has led to debate over whether CNVs are mutations that propagate disease or whether they are a normal, and even essential, aspect of cell biology. We show that 1,000N polyploid trophoblast giant cells (TGCs) of the mouse placenta contain 47 regions, totaling 138 Megabases, where genomic copies are underrepresented (UR). UR domains originate from a subset of late-replicating heterochromatic regions containing gene deserts and genes involved in cell adhesion and neurogenesis. While lineage-specific CNVs have been identified in mammalian cells, classically in the immune system where V(D)J recombination occurs, we demonstrate that CNVs form during gestation in the placenta by an underreplication mechanism, not by recombination nor deletion. Our results reveal that large scale CNVs are a normal feature of the mammalian placental genome, which are regulated systematically during embryogenesis and are propagated by a mechanism of underreplication.