Citations

Abstract Introduction: Urine cell-free DNA (cfDNA) is a promising ultra-noninvasive analyte for genitourinary cancers (GU) cancers. In this study, we investigated urine cfDNA fragmentomic features and evaluated their ability to predict the tumor tissue of origin. Methods: A total of 204 GU cancer patients and 34 healthy adults were enrolled in this prospective study. Preoperative urine was collected from 89 bladder cancer (BC) patients (64% muscle-invasive) undergoing cystectomy, 65 renal cell carcinoma (RCC) patients undergoing nephrectomy and 50 metastatic prostate cancer (mPC) patients. Urine cfDNA was isolated and sequenced at 5x genome-wide coverage on a NovaSeq S4 flow cell. To identify malignant tissue of origin, we analyzed somatic copy number alterations (CNAs), genome-wide fragment length ratios, and 4-mer end motifs. CNAs and tumor fraction (TFx) were quantified using ichorCNA in 1-Mb windows, while chromosome arm-level fragment count z-scores and fragment length ratios were calculated in 5-Mb bins across the genome. Relative frequencies of all 256 possible 4-mer end motifs were calculated for each sample. Individual logistic regression models were developed based on each fragmentomic feature using leave-one-out cross validation (LOOCV). Outputs from these models were combined into multi-feature XGBoost models. Cohort-specific end-motif enrichments were identified using Kruskal-Wallis tests with FDR correction. Motifs were then hierarchically clustered based on their cohort-wise median frequencies to define groups of motifs with shared fragmentation patterns. Results: Across urine cfDNA in the GU cohort, BC showed the highest mean TFx at 10.0%, followed by mPC at 3.6% and RCC at 2.7%. A machine learning model based on 4-mer end motifs achieved a mean ROC AUC of 0.933 with classification accuracies of 63% for healthy, 87% for BC, 72% for mPC, and 83% for RCC. This model outperformed models based on ichorCNA (AUC = 0.767), chromosome arm fragment-count z-scores (AUC = 0.827), and fragment length ratios (AUC = 0.674). Combining end motifs with other fragmentomic features did not improve prediction accuracy. Multi-feature models incorporating end motifs with ichorCNA, chromosome arm-level z-scores, or fragment-length ratios yielded mean AUCs between 0.908-0.924, lower than the end-motif only model. Kruskal-Wallis tests identified 179 of 256 end motifs (79%) as significantly different across healthy, BC, mPC, and RCC. Hierarchical clustering of cohort-level median motifs revealed distinct end motif patterns: Motifs beginning with CT were enriched in RCC; AAAA was strongly enriched in BC; and motifs beginning with CA, AC, and AG were enriched in mPC. Conclusion: Urine cfDNA 4-mer end motifs capture GU cancer-specific biological patterns that enable accurate tissue-of-origin prediction, underscoring their promise as a diagnostic tool. Citation Format: Jessica Linford, Pradeep S. Chauhan, Irfan Alahi, Yohan Kim, Arpit Panda, Nathan Colon, Ryan Mueller, Faridi Qaium, Eric H. Kim, Melissa A. Reimers, Zachary L. Smith, Woodson W. Smelser, Fabrice Lucien-Matteoni, Aadel A. Chaudhuri. Urine cfDNA 4-mer end-motif signatures outperform other fragmentomic features for tissue-of-origin classification in genitourinary cancers [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 2590.

Abstract Background: Liquid biopsy approaches for cancer detection often rely on tissue-specific methylation or fragmentation features to infer tumor origin. However, a pan-aerodigestive cancer detection strategy is clinically valuable given shared risk exposures, field cancerization, and frequent diagnostic ambiguity among lung, esophageal, and head & neck cancers. We hypothesized that advanced-stage disease exhibits convergent chromatin remodeling, producing a dominant malignancy-associated fragmentomic signal that supersedes subtler tissue-of-origin patterns. Methods: Whole-genome EM-seq (∼15×) was performed on plasma cfDNA from 120 patients with stage III/IV aerodigestive cancers (lung, esophageal, head & neck; squamous and adenocarcinoma) and 60 age-matched controls. Four fragmentomic modalities were assessed: (1) genome-wide fragmentation in 5-Mb bins (DELFI-like), (2) NMF spectral components, (3) 4-mer end motifs, and (4) promoter-centric nucleosome scores (Griffin). All preprocessing occurred strictly within each training fold of a repeated stratified 10-fold CV (100 repeats) to prevent data leakage. No test-sample information was used in training. Hyperparameters were tuned only within training partitions. Random Forest classifiers were trained under this fully nested scheme. Results: The multimodal classifier robustly detected advanced aerodigestive malignancies with a mean AUROC 0.86 and AUPRC 0.94. However, modality dissection revealed a clear performance hierarchy: nucleosome positioning (Griffin) alone achieved AUROC 0.85 and AUPRC 0.93, nearly matching the full integrated model, whereas end motifs performed poorly (AUROC 0.61). Biological interpretation demonstrated prominent nucleosome depletion centered at promoters of proliferation-associated genes, indicating a convergent epigenetic state of promoter opening across late-stage aerodigestive malignancies. Conclusions: Advanced aerodigestive cancers exhibit a dominant, convergent chromatin remodeling signature detectable in plasma cfDNA, driven primarily by altered nucleosome positioning. In late-stage disease, this shared malignancy-associated fragmentomic signal outweighs tissue-of-origin-dependent patterns. These findings highlight the utility of nucleosome-based fragmentomics not merely for detection, but for resolving diagnostic ambiguity in patients with indeterminate masses and for monitoring total tumor burden in high-risk aerodigestive malignancies. Citation Format: Axel Misael Hidalgo, Ayesha Hashmi, Jeff Szymanski, Pradeep Chauhan, Lilli Greiner, Faridi Qaium, Daniel J. Ma, David M. Routman, Katie M. Van Abel, Aadel A. Chaudhuri.. Convergent chromatin remodeling enables pan-aerodigestive detection of advanced malignancies in plasma cfDNA [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 3862.

Abstract Introduction: Long intergenic noncoding RNAs (lincRNAs) and circular (circ) RNAs have been shown to play key roles in carcinogenesis and tumor progression. Urine represents a promising source of tumor-reflective signals shed directly from the urothelial microenvironment in bladder cancer (BC) patients. Here, we evaluate the urine cell-free (cf) RNA transcriptome in a cohort of 50 patients with localized BC. Methods: A total of 50 localized BC patients and 24 healthy adults were enrolled. Pre-operative urine (collected on the day of radical cystectomy) was processed to isolate cfRNA from urine supernatant. Libraries were prepared using an Illumina tagmentation-based RNA enrichment workflow and sequenced to 100M paired-end reads per sample on a NovaSeq X. FASTQ files were concatenated per sample and QC-checked with RNA-SeQC and Samtools. Reads were aligned to GRCh38/hg38 with STAR, and gene counts were generated using Feature Counts. Linear differential expression analysis (DEA) was performed with DESeq2. For discovery we used random halves of TCGA-BLCA and healthy urine cfRNA samples, alongside a second DEA with random halves of urine BC cfRNA vs. controls; overlapping upregulated genes were validated in the remaining samples. For circular (circ) RNAs, annotations from MiOncoCirc were used; junctions with ≥5 supporting reads were retained, and DEA was conducted with edgeR. Results: In this study, the median age was 68 years, and 80% (40/50) were male. Among the cohort, 66% (33/50) had muscle-invasive bladder cancer (MIBC), of whom 54% received neoadjuvant chemotherapy. Differential expression analysis of TCGA-BLCA tumors and BC urine cfRNA, identified 27 genes commonly upregulated, with 23 of the 27 shared genes strikingly being long intergenic noncoding RNAs (lincRNAs). These BC/urine-specific lincRNAs—including LINC00930, LINC01004, DLG5-AS1, APTR, PSORS1C3, and SERPINB9P1—have known roles in proliferation, EMT, metabolism, and immune signaling. Outside of lincRNAs, tumor-reflective coding transcripts in BC/urine were enriched for autophagy, driven by MAP1LC3B2.TCGA-BLCA survival analysis showed that LINC00930 and PSORS1C3 were associated with worse overall survival. Expanding beyond linear RNA sequences, analysis of urine circRNAs from BC patients revealed a strong cancer-specific dysregulation pattern, with 77 circRNAs significantly upregulated compared with healthy controls. Upregulated urine circRNAs arose from host genes involved in BC-relevant pathways, including invasion (DOCK1, ARHGAP5), autophagy/ER stress (VMP1, ERN1), and luminal differentiation (GRHL2, ESRP1). Conclusion: Cell-free RNA profiling reveals a distinct transcriptomic signature including differential lincRNA- and circRNA-driven autophagy, EMT, proliferation, metabolic, differentiation, and immune signaling programs in the urine of bladder cancer patients. Citation Format: Pradeep Chauhan, Li Lin, Irfan Alahi, Jessica Linford, Lilli J. Greiner, Ayesha Hashmi, Anushka Viswanathan, Nathan Colon, Faridi Qaium, Peter K. Harris, John Sheng, Jacob J. Orme, Eric H. Kim, Zachary L. Smith, Woodson Smelser, Christopher A. Maher, Aadel A. Chaudhuri. Urine cell-free RNA captures tumor-reflective long intergenic noncoding and circular RNA programs in bladder 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 1133.

Abstract Background: Spatial cellular ecosystems in the tumor microenvironment (TME) form dynamic signaling hubs that critically influence cancer disease progression and response to therapy, including response to immune checkpoint inhibitors (ICIs). However, tumor sampling bias and the impracticality of acquiring serial tumor biopsies have made it challenging to profile the TME clinically. To address this challenge, we developed Liquid EcoTyper, a multi-analyte AI framework to quantify TME spatial cellular ecosystems—termed spatial ecotypes (SEs)—noninvasively from cell-free DNA (cfDNA) methylation data. Methods: In previous work, we discovered nine spatial ecotypes, including SEs associated with ICI response (e.g., SE7 with benefit, SE4 with resistance), and quantified them from 1,249 bulk tumor RNA-seq profiles of melanoma, non-small cell lung cancer (NSCLC), and bladder cancer (BC) patients (Cancer Res (2025) 85 (8_Supplement_1): 153). Here we trained Liquid EcoTyper, an interpretable deep learning model, to transfer SE profiling to cell-free DNA using CpG methylation profiles. Performance was quantified using simulated and real data, including methylation profiles (EM-seq) of plasma cfDNA paired with tumor biopsy-derived SE levels determined by EM-seq (n=20 pairs) or 10x Visium (n=15 pairs). Clinical outcome prediction was evaluated using pretreatment plasma from patients treated with ICI alone (n=78 melanoma pts), chemotherapy and ICI (n=25 NSCLC pts), and enfortumab vedotin and ICI (n=10 localized muscle-invasive BC (MIBC) pts). Tumor mutational burden (TMB), tumor PD-L1, or circulating tumor DNA (ctDNA) were profiled as a comparator where possible. Results: We observed striking concordance between plasma-derived SE levels and tumor biopsy-confirmed SE levels from the same patients, demonstrating that liquid biopsy analysis of cfDNA methylomes can recapitulate TME spatial biology. Underscoring specificity, no reliable SE signal was detectable from peripheral blood mononuclear cells. In pre-ICI plasma, liquid SE levels were significantly associated with ICI-based response (SE7) versus resistance (SE4) in patients with melanoma, NSCLC, and MIBC, with a mean cross-cancer area under the curve (AUC) of 0.87 for SE7 (range 0.8-0.97) and 0.81 for SE4 (range 0.76-0.85)—far surpassing TMB, tumor PD-L1, and ctDNA in evaluable patients, both in binary analyses of response and multivariable models of survival. Our results also showed potential to predict pathologic complete response in the neoadjuvant setting. Conclusion: Liquid EcoTyper is a scalable noninvasive AI-based framework for spatiotemporal assessment of the TME. By enabling TME-informed risk stratification without surgical biopsies, Liquid EcoTyper has potential to enable more precise and actionable decision making in patients with solid tumors. Citation Format: Erin L. Brown, Wubing Zhang, Abul Usmani, Noah Earland, Ayesha Hashmi, Chibuzor Olelewe, Anushka Viswanathan, Pradeep S. Chauhan, Minji Kang, Chloé B. Steen, Hyun Soo Jeon, Susanna Avagyan, Irfan Alahi, Nicholas P. Semenkovich, Matteo Bergsagel, Janella C. Schwab, Chloe M. Sachs, Faridi Qaium, Peter K. Harris, Antonella Bacchiocchi, Qingyuan Cai, Andrew J. Gentles, Rondell P. Graham, Peter C. Lucas, Ryan C. Fields, Jacob J. Orme, Aaron S. Mansfield, Mario Sznol, Ruth Halaban, David Y. Chen, Aaron M. Newman, Aadel A. Chaudhuri. Liquid biopsy profiling of the tumor microenvironment to determine response to immunotherapy regimens across solid tumors [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 94.

Abstract We developed and cross‑validated an integrated circulating tumor DNA assay incorporating SNVs, indels, CNAs, and SVs to distinguish neurofibromatosis type 1 patients with malignant peripheral nerve sheath tumors from those with benign plexiform neurofibromas or tumor‑free controls. Among 82 participants, the assay achieved an AUC of 0.904, compared with 0.735 for genome‑wide CNAs. We also detected recurrent disease‑specific driver alterations, relapse up to three months before diagnosis, and early clearance consistent with durable remission.

Abstract Introduction: We previously developed a urine liquid biopsy assay to measure urine tumor DNA (utDNA) levels in bladder cancer (BC) patients. Here, we extended this approach significantly by layering in cell-free DNA fragmentomic and cell-free RNA analysis, and extending the method to the full spectrum of genitourinary (GU) cancers, to enable both GU cancer detection and tumor tissue of origin prediction. Methods: A total of 155 GU cancer patients and 34 healthy adults were enrolled in this prospective study. Preoperative urine was collected from 94 bladder cancer (BC) patients, including 65 with muscle-invasive disease, all undergoing cystectomy. Additionally, urine samples were collected from 18 renal cell carcinoma (RCC) patients undergoing nephrectomy and 42 metastatic prostate cancer (mPC) patients. Urine cfDNA and cfRNA were isolated and sequenced, with cfRNA sequenced to 100 million reads per sample on an Illumina NovaSeq X and cfDNA sequenced at 5x genome-wide coverage on a Novoseq S4 flow cell. To identify malignant tissue of origin, we analyzed somatic copy number alterations (CNAs), genome-wide fragment length ratios, and 4-mer end motifs. Tumor fraction (TFx) was estimated using ichorCNA, while DELFI was employed for fragment length ratios in 5 megabase bins across the genome and chromosome arm-level z-scores. Individual logistic regression models were developed based on each fragmentomic feature, and validated using a training and held-out testing approach. Outputs from these models were combined into a single comprehensive model, also trained and held-out tested using the same approach. Urine cfRNA transcript isoform abundance was quantified using Salmon and normalized to Transcripts Per Million (TPM) to account for sequencing depth variations. Results: Bladder cancer (BC) showed the highest mean tumor fraction (TFx, 10.5%), followed by RCC (3.8%) and mPC (3.4%). Among 94 BC patients, 42% achieved a pCR, while 58% had residual disease and did not. Patients without pCR had significantly higher urine TFx compared to those with pCR (14.7% vs. 4.7%, p=0.0008). A machine learning model based on genome-wide fragment length ratios achieved an ROC AUC of 0.74, while models using genome-wide CNAs and end motifs achieved AUCs of 0.81 and 0.88, respectively. Combining all features into a single model yielded an AUC of 0.92, with classification accuracies of 89% for bladder cancer, 80% for RCC, and 60% for prostate cancer. Additionally, urine cfRNA analysis using the top 1, 000 coding transcripts achieved an AUC of 0.95, with classification accuracies of 100% for mPC, 87% for BC, and 78% for RCC. Conclusion: Fragmentomic features from urine cfDNA and transcriptomic features from urine cfRNA demonstrate strong potential for predicting tumor tissue-of-origin in GU cancers, underscoring the promise of non-invasive diagnosis across the spectrum of GU malignancies through urine liquid biopsy. Citation Format: Pradeep Singh Chauhan, Irfan Alahi, Jessica Linford, Arpit Panda, Nathan Colon, Ryan Mueller, Faridi Qaium, Peter K. Harris, John Sheng, Eric H. Kim, Melissa A. Reimers, Zachary L. Smith, Woodson W. Smelser, Aadel A. Chaudhuri. Predicting genitourinary cancer tissue-of-origin using urine cell-free DNA fragmentomics and cell-free RNA [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 3258.

Background: Multicellular programs in the tumor microenvironment (TME) drive cancer pathogenesis and response to therapy but remain challenging to identify and profile clinically. Here, we present Spatial EcoTyper, a multimodal machine learning framework for large-scale profiling of spatially dependent cell states and multicellular ecosystems, termed spatial ecotypes (SEs). Methods and Findings: By integrating five million single-cell and spot-level spatial transcriptomes from 10 diverse human neoplasms, including carcinomas and melanomas, we identified a rich tapestry of conserved regional plasticity across nine TME cell types. We then applied Spatial EcoTyper to deeply dissect these data, revealing nine novel SEs with broad conservation - each with unique biology, TME cell states, geospatial features, and clinical outcome associations. SEs were linked to specific biogeographic domains, including the tumor core, margin, and adjacent stroma, and were remarkably well-validated in 235 held-out tumor samples profiled by either single-cell RNA sequencing, sequencing-based spatial transcriptomics (10x Visium), or high-resolution imaging-based spatial transcriptomics (Vizgen MERSCOPE and NanoString CosMx). By deconvolving SEs from over 1,200 bulk tumor RNA-seq profiles, we discovered striking associations between SE levels and response to immune checkpoint inhibitors (ICIs), outperforming nearly 40 potential correlates of ICI response, including previously described signatures of tertiary lymphoid structures and tumor tissue PD-L1 expression. Notably, baseline levels of a proinflammatory SE - localized to the tumor core and defined by a hub of unique T cell, macrophage, fibroblast, and endothelial cell states - strongly forecasted ICI benefit, while another SE - localized to the tumor margin and defined by myofibroblast and hypoxia-associated endothelial cell states - portended ICI resistance. Conclusion: Our data reveal fundamental units of TME organization and demonstrate a platform for large-scale profiling of spatial cellular ecosystems in any tissue, with implications for improved forecasting of immunotherapy response and the development of novel therapeutic strategies. Citation Format: Wubing Zhang,Erin L. Brown,Abul Usmani,Noah Earland,Hyun Soo Jeon,Irfan Alahi,Nicholas P. Semenkovich,Janella C. Schwab,Chloé B. Steen,Chloe M. Sachs,Bilge Gungoren,Faridi Qaium,Peter K. Harris,Jacqueline Mudd,Qingyuan Cai,Antonella Bacchiocchi,Ryan C. Fields,Mario Sznol,Ruth Halaban,David Y. Chen,Aadel A. Chaudhuri,Aaron M. Newman. High-resolution profiling of the tumor microenvironment with spatial ecotypes [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 153.

Abstract Purpose: Metastatic castration-resistant prostate cancer (mCRPC) resistant to androgen receptor signaling inhibitors (ARSI) is often lethal. Liquid biopsy biomarkers for this deadly form of disease remain under investigation, and underpinning mechanisms remain ill-understood. Experimental Design: We applied targeted cell-free DNA (cfDNA) sequencing to 126 patients with mCRPC from three academic cancer centers and separately performed genome-wide cfDNA methylation sequencing on 43 plasma samples collected prior to the initiation of first-line ARSI treatment. To analyze the genome-wide sequencing data, we performed nucleosome positioning and differential methylated region analysis. We additionally analyzed single-cell and bulk RNA sequencing data from 14 and 80 patients with mCRPC, respectively, to develop and validate a stem-like signature, which we inferred from cfDNA. Results: Targeted cfDNA sequencing detected AR/enhancer alterations prior to first-line ARSIs that correlated with significantly worse progression-free survival (P = 0.01; HR = 2.12) and overall survival (P = 0.02; HR = 2.48). Plasma methylome analysis revealed that AR/enhancer lethal mCRPC patients have significantly higher promoter-level hypomethylation than AR/enhancer wild-type mCRPC patients (P < 0.0001). Moreover, gene ontology and CytoTRACE analysis of nucleosomally more accessible transcription factors in cfDNA revealed enrichment for stemness-associated transcription factors in patients with lethal mCRPC. The resulting stemness signature was then validated in a completely held-out cohort of 80 patients with mCRPC profiled by tumor RNA sequencing. Conclusions: We analyzed a total of 220 patients with mCRPC, validated the importance of cell-free AR/enhancer alterations as a prognostic biomarker in lethal mCRPC, and showed that the underlying mechanism for lethality involves reprogramming developmental states toward increased stemness. See related commentary by Nawfal et al., p. 7

Abstract Purpose: Early detection of neurofibromatosis type 1 (NF1)–associated peripheral nerve sheath tumors (PNST) informs clinical decision-making, enabling early definitive treatment and potentially averting deadly outcomes. In this study, we describe a cell-free DNA (cfDNA) fragmentomic approach that distinguishes nonmalignant, premalignant, and malignant forms of PNST in the cancer predisposition syndrome, NF1. Experimental Design: cfDNA was isolated from plasma samples of a novel cohort of 101 patients with NF1 and 21 healthy controls and underwent whole-genome sequencing. We investigated diagnosis-specific signatures of copy-number alterations with in silico size selection as well as fragment profiles. Fragmentomics were analyzed using complementary feature types: bin-wise fragment size ratios, end motifs, and fragment non-negative matrix factorization signatures. Results: The novel cohort of patients with NF1 validated that our previous cfDNA copy-number alteration–based approach identifies malignant PNST (MPNST) but cannot distinguish between benign and premalignant states. Fragmentomic methods were able to differentiate premalignant states including atypical neurofibromas (AN). Fragmentomics also adjudicated AN cases suspicious for MPNST, correctly diagnosing samples noninvasively, which could have informed clinical management. Conclusions: Novel cfDNA fragmentomic signatures distinguish AN from benign plexiform neurofibromas and MPNST, enabling more precise clinical diagnosis and management. This study pioneers the early detection of malignant and premalignant PNST in NF1 and provides a blueprint for decentralizing noninvasive cancer surveillance in hereditary cancer predisposition syndromes.

Abstract Background: Metastatic castration-resistant prostate cancer (mCRPC) resistant to androgen receptor (AR)-targeted agents is often lethal. Unfortunately, biomarkers for this deadly disease remain under investigation, and underpinning mechanisms are ill-understood. Here, we applied epigenomic approaches to identify cell-free (cf) DNA features associated with these lethal AR-resistant mCRPC patients. Methods: Plasma from 99 mCRPC patients was collected from two independent institutions, either prior to first-line AR-targeted therapy (n = 63) or during treatment (n = 36). We applied EnhanceAR-Seq to detect alterations in the AR locus including the upstream enhancer. We then applied genome-wide cfDNA epigenomics to 43 pre-treatment plasma samples and performed both fragmentomics and methylation sequencing to delineate the biology of AR/enhancer-altered lethal mCRPC. Finally, we identified stemness genes by applying CytoTRACE to external single-cell RNA-seq data from mCRPC, and applied these to cfDNA via promoter-level methylation analysis. We then assessed these stem-associated genes in tumor bulk RNA-seq data from an independent mCRPC cohort. Results: AR/enhancer alterations were detected in 44% of pre-treatment cfDNA samples and correlated with significantly worse progression-free survival (PFS) (HR = 2.12, P = 0.01) and overall survival (OS) (HR = 2.48; P = 0.02). AR/enhancer alterations were detected in 19% of on-treatment cfDNA samples and were also associated with profoundly worse PFS (HR = 15.38, P = 0.0003) and OS (HR = 15.53, P = 0.002). Further, genome-wide cfDNA fragmentomic analysis revealed that binding sites for developmentally relevant transcription factors such as FOXA1, NKX-3 and HOXB13 were significantly more accessible in AR/enhancer-altered lethal mCRPC patients. Gene set enrichment analysis of the 20 most accessible transcription factors revealed several pathways associated with stem cell development. Interestingly, the 10 most stem-associated genes obtained using CytoTRACE applied to an external cohort of single-cell RNA-seq data from mCRPC (He et al. 2021) were promoter-hypomethylated in the plasma cfDNA of AR/enhancer-altered lethal mCRPC patients (P = 0.005). Further, metagene analysis of these stem-associated genes revealed worse PFS by Kaplan-Meier analysis (P = 0.04), which was confirmed by multivariate cox proportional hazards modeling. Our stemness metagene further stratified overall survival (P = 0.03) when applied to an external cohort of 80 mCRPC patients from Abida et al. 2019. Conclusions: Alterations in AR including the upstream enhancer are remarkably prognostic in mCRPC with pre-treatment predictive potential. The mechanism can be inferred via cfDNA epigenomic analysis and appears to involve increased tumor cell stemness in AR-altered lethal mCRPC. Citation Format: Pradeep Singh Chauhan, Irfan Alahi, Savar Sinha, Ryan Mueller, Alexander L. Shiang, Jace Webster, Ha X. Dang, Debanjan Saha, Lilli Greiner, Breanna Yang, Elisa M. Ledet, Ramandeep K. Babbra, Wenjia Feng, Peter K. Harris, Faridi Qaium, Ellen B. Jaeger, Patrick J. Miller, Sydney A. Caputo, Oliver A. Sartor, Russell K. Pachynski, Christopher A. Maher, Aadel A. Chaudhuri. Epigenomic analysis of plasma cell-free DNA identifies stemness features associated with worse prognosis in AR-altered lethal metastatic castration resistant prostate cancer [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2024; Part 1 (Regular Abstracts); 2024 Apr 5-10; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2024;84(6_Suppl):Abstract nr 983.