Biology subjects

Berger, M. F.

Publications and source records attributed to Berger, M. F..

8 recordsLinked to original sources

bioRxiv2026

Generalist large language models complement tailor-made predictors for tumor genomics interpretation

General-purpose large language models (LLMs) are trained on large corpora to acquire broad knowledge, but whether LLMs can replace, or augment, task-specific models is unclear. We evaluated LLMs on three real-world, clinically important tumor genomic interpretation tasks, in order of increasing difficulty: (i) distinguishing tumor from non-tumor mutations (n=34,415 variants), (ii) distinguishing driver from passenger mutations (n=13,469 variants), and (iii) inferring cancer type from tumor sequencing reports across multiple assays and institutions (n=102,791 samples). The best general-purpose LLMs performed as well as the benchmark tailor-made predictor for task (i). Ensembling tailor-made models with zero-shot LLMs improved their performance for tasks (i) and (ii). For task (iii), LLMs outperformed or supplemented tailor-made models on out-of-distribution data. Without fine-tuning, current LLMs already can be useful in clinical genomic interpretation by adding complementary expertise to tailor-made, state-of-the-art predictors.

genomics↗

bioRxiv2024

Tracking clonal evolution of drug resistance in ovarian cancer patients by exploiting structural variants in cfDNA

Drug resistance is the major cause of therapeutic failure in high-grade serous ovarian cancer (HGSOC). Yet, the mechanisms by which tumors evolve to drug resistant states remains largely unknown. To address this, we aimed to exploit clone-specific genomic structural variations by combining scaled single-cell whole genome sequencing with longitudinally collected cell-free DNA (cfDNA), enabling clonal tracking before, during and after treatment. We developed a cfDNA hybrid capture, deep sequencing approach based on leveraging clone-specific structural variants as endogenous barcodes, with orders of magnitude lower error rates than single nucleotide variants in ctDNA (circulating tumor DNA) detection, demonstrated on 19 patients at baseline. We then applied this to monitor and model clonal evolution over several years in ten HGSOC patients treated with systemic therapy from diagnosis through recurrence. We found drug resistance to be polyclonal in most cases, but frequently dominated by a single high-fitness and expanding clone, reducing clonal diversity in the relapsed disease state in most patients. Drug-resistant clones frequently displayed notable genomic features, including high-level amplifications of oncogenes such as CCNE1, RAB25, NOTCH3, and ERBB2. Using a population genetics Wright-Fisher model, we found evolutionary trajectories of these features were consistent with drug-induced positive selection. In select cases, these alterations impacted selection of secondary lines of therapy with positive patient outcomes. For cases with matched single-cell RNA sequencing data, pre-existing and genomically encoded phenotypic states such as upregulation of EMT and VEGF were linked to drug resistance. Together, our findings indicate that drug resistant states in HGSOC pre-exist at diagnosis and lead to dramatic clonal expansions that alter clonal composition at the time of relapse. We suggest that combining tumor single cell sequencing with cfDNA enables clonal tracking in patients and harbors potential for evolution-informed adaptive treatment decisions.

cancer biology↗

bioRxiv2024

Machine learning predictions improve identification of real-world cancer driver mutations

Characterizing and validating which mutations influence development of cancer is challenging. Machine learning has delivered significant advances in protein structure prediction, but its utility for identifying cancer drivers is less explored. We evaluated multiple computational methods for identifying cancer driver alterations. For identifying known drivers, methods incorporating protein structure or functional genomic data outperformed methods trained only on evolutionary data. We further validated VUSs annotated as pathogenic by testing their association with overall survival in two cohorts of patients with non-small cell lung cancer (N=7,965 and 977). "Pathogenic" VUSs in KEAP1 and SMARCA4 identified by several methods were associated with worse survival, unlike "benign" VUSs. "Pathogenic" VUSs exhibited mutual exclusivity with known oncogenic alterations at the pathway level, further suggesting biological validity. Despite training primarily on germline, rather than somatic, mutation data, computational predictions contribute to a more comprehensive understanding of tumor genetics as validated by real-world data.

genomics↗

bioRxiv2022

Machine learning guided signal enrichment for ultrasensitive plasma tumor burden monitoring

In solid tumor oncology, circulating tumor DNA (ctDNA) is poised to transform care through accurate assessment of minimal residual disease (MRD) and therapeutic response monitoring. To overcome the sparsity of ctDNA fragments in low tumor fraction (TF) settings and increase MRD sensitivity, we previously leveraged genome-wide mutational integration through plasma whole genome sequencing (WGS). We now introduce MRD-EDGE, a composite machine learning-guided WGS ctDNA single nucleotide variant (SNV) and copy number variant (CNV) detection platform designed to increase signal enrichment. MRD-EDGE uses deep learning and a ctDNA-specific feature space to increase SNV signal to noise enrichment in WGS by 300X compared to our previous noise suppression platform MRDetect. MRD-EDGE also reduces the degree of aneuploidy needed for ultrasensitive CNV detection through WGS from 1Gb to 200Mb, thereby expanding its applicability to a wider range of solid tumors. We harness the improved performance to track changes in tumor burden in response to neoadjuvant immunotherapy in non-small cell lung cancer and demonstrate ctDNA shedding in precancerous colorectal adenomas. Finally, the radical signal to noise enrichment in MRD-EDGE enables de novo mutation calling in melanoma without matched tumor, yielding clinically informative TF monitoring for patients on immune checkpoint inhibition.

genomics↗

bioRxiv2021

ERα-LBD, a novel isoform of estrogen receptor alpha, promotes breast cancer proliferation and endocrine resistance

Estrogen receptor alpha (ER) drives mammary gland development and breast cancer (BC) growth through an evolutionarily conserved linkage of DNA binding and hormone activation functions. Therapeutic targeting of the hormone binding pocket is a widely utilized and successful strategy for breast cancer prevention and treatment. However, resistance to this endocrine therapy is frequently encountered and may occur through bypass or reactivation of ER-regulated transcriptional programs. We now identify the induction of a novel ER isoform, ER-LBD, that is encoded by an alternative ESR1 transcript and lacks the activation function and DNA binding domains. Despite lacking the transcriptional activity, ER-LBD is found to promote breast cancer growth and resistance to the ER antagonist fulvestrant. ER-LBD is predominantly localized to the cytoplasm and mitochondria of BC cells and leads to enhanced glycolysis, respiration and stem-like features. Intriguingly, ER-LBD expression and function does not appear to be restricted to cancers that express full length ER but also promotes growth of triple negative breast cancers and ER-LBD transcript (ESR1-LBD) is also present in BC samples from both ER(+) and ER(-) human tumors. These findings point to ER-LBD as a potential mediator of breast cancer progression and therapy resistance. SIGNIFICANCE STATEMENTEndocrine resistant and metastatic breast cancer (BC) is a clinically significant problem. Our study of fulvestrant resistant cancer cells led to the discovery of a novel ER isoform which we call ER-LBD. Encoded by a truncated transcript variant (ESR1-LBD) and lacking the N-terminal domains (activation of transcription and DNA binding), ER-LBD displays a unique role in BC tumorigenesis and progression by mechanisms that may involve metabolic and cell growth advantages, stemness and therapy resistance. Importantly, ESR1-LBD is preferentially expressed in human breast tumor tissues and may be used as prognostic marker in BC.

cancer biology↗

bioRxiv2020

Chromatin accessibility profiles of castration-resistant prostate cancers reveal novel subtypes and therapeutic vulnerabilities

In castration-resistant prostate cancer (CRPC), the loss of androgen receptor (AR)-dependence due to lineage plasticity, which has become more prevalent, leads to clinically highly aggressive tumors with few therapeutic options and is mechanistically poorly defined. To identify the master transcription factors (TFs) of CRPC in a subtype-specific manner, we derived and collected 29 metastatic human prostate cancer organoids and cell lines, and generated ATAC-seq, RNA-seq and DNA sequencing data. We identified four subtypes and their master TFs using novel computational algorithms: AR-dependent; Wnt-dependent, driven by TCF; neuroendocrine, driven by ASCL1 and NEUROD1 and stem cell-like (SCL), driven by the AP-1 family. The transcriptomic signatures of these four subtypes enabled the classification of 370 patients. We find that AP-1 co-operates with the inhibitable YAP/TAZ/TEAD pathway in the SCL subtype, the second most common group of CRPC tumors after AR-dependent. Together, this molecular classification reveals new drug targets and can potentially guide therapeutic decisions.

cancer biology↗

bioRxiv2020

MiMSI - a deep multiple instance learning framework improves microsatellite instability detection from tumor next-generation sequencing

Microsatellite instability (MSI) is a critical phenotype of cancer genomes and an FDA-recognized biomarker that can guide treatment with immune checkpoint inhibitors. Recent work has demonstrated that next-generation sequencing data can be used to identify samples with MSI-high phenotype. However, low tumor purity, as frequently observed in routine clinical samples, poses a challenge to the sensitivity of existing algorithms. To overcome this critical issue, we developed MiMSI, an MSI classifier based on deep neural networks and trained using a dataset that included low tumor purity MSI cases in a multiple instance learning framework. On a challenging yet representative set of cases, MiMSI showed higher sensitivity (0.940) and auROC (0.988) than MSISensor(sensitivity: 0.57; auROC: 0.911), an open-source software previously validated for clinical use at our institution using MSK-IMPACT large panel targeted NGS data.

bioinformatics↗

bioRxiv2019

A rectal cancer model establishes a platform to study individual responses to chemoradiation

Rectal cancer (RC) is a challenging disease to treat that requires chemotherapy, radiation, and surgery to optimize outcomes for individual patients. No accurate model of RC exists to answer fundamental research questions relevant to individual patients. We established a biorepository of 32 patient-derived RC organoid cultures (tumoroids) from patients with primary, metastatic, or recurrent disease. RC tumoroids retained molecular features of the tumors from which they were derived, and their ex vivo responses to clinically relevant chemotherapy and radiation treatment correlate well with responses noted in individual patients tumors. Upon engraftment into murine rectal mucosa, human RC tumoroids gave rise to invasive rectal cancer followed by metastasis to lung and liver. Importantly, engrafted tumors closely reflected the heterogenous sensitivity to chemotherapy observed clinically. Thus, the biology and drug sensitivity of RC clinical isolates can be efficiently interrogated using an organoid-based, in vitro platform coupled with endoluminal propagation in animals.

cancer biology↗