Stefan G. Stark
A5038601102- Cancer Genomics and Diagnostics
- PI3K/AKT/mTOR signaling in cancer
- Bioinformatics and Genomic Networks
- Single-cell and spatial transcriptomics
- RNA modifications and cancer
- RNA Research and Splicing
- Protein Kinase Regulation and GTPase Signaling
- Gene expression and cancer classification
- Genomics and Phylogenetic Studies
- Cancer-related molecular mechanisms research
- Biochemical and Molecular Research
- Genomics and Chromatin Dynamics
- Evolution and Genetic Dynamics
- Genetic factors in colorectal cancer
- Chromosomal and Genetic Variations
- Epigenetics and DNA Methylation
- Cell Image Analysis Techniques
- Nutrition, Genetics, and Disease
- Bayesian Methods and Mixture Models
- Genomics and Rare Diseases
- Pancreatic and Hepatic Oncology Research
- Complex Network Analysis Techniques
- Biomedical Text Mining and Ontologies
- Telomeres, Telomerase, and Senescence
- DNA Repair Mechanisms
ETH Zurich
2017-2024
SIB Swiss Institute of Bioinformatics
2017-2024
Memorial Sloan Kettering Cancer Center
2015-2023
University Hospital of Zurich
2017-2023
University of Zurich
2021-2022
Alexandru Ioan Cuza University
2022
Korea University
2020
Life Science Zurich
2020
German Cancer Research Center
2013
Heidelberg University
2013
Our comprehensive analysis of alternative splicing across 32 The Cancer Genome Atlas cancer types from 8,705 patients detects events and tumor variants by reanalyzing RNA whole-exome sequencing data. Tumors have up to 30% more than normal samples. Association somatic with confirmed known trans associations in SF3B1 U2AF1 identified additional trans-acting (e.g., TADA1, PPP2R1A). Many tumors thousands not detectable samples; on average, we ≈930 exon-exon junctions ("neojunctions") typically...
Abstract Transcript alterations often result from somatic changes in cancer genomes 1 . Various forms of RNA have been described cancer, including overexpression 2 , altered splicing 3 and gene fusions 4 ; however, it is difficult to attribute these underlying genomic owing heterogeneity among patients tumour types, the relatively small cohorts for whom samples analysed by both transcriptome whole-genome sequencing. Here we present, our knowledge, most comprehensive catalogue...
Abstract Understanding and predicting molecular responses in single cells upon chemical, genetic or mechanical perturbations is a core question biology. Obtaining single-cell measurements typically requires the to be destroyed. This makes learning heterogeneous perturbation challenging as we only observe unpaired distributions of perturbed non-perturbed cells. Here leverage theory optimal transport recent advent input convex neural architectures present CellOT, framework for response...
Abstract The impact of somatic structural variants (SVs) on gene expression in cancer is largely unknown. Here, as part the ICGC/TCGA Pan-Cancer Analysis Whole Genomes (PCAWG) Consortium, which aggregated whole-genome sequencing data and RNA from a common set 1220 cases, we report hundreds genes for presence within 100 kb an SV breakpoint associates with altered expression. For majority these genes, increases rather than decreases corresponding events. Up-regulated cancer-associated impacted...
Recent technological advances have led to an increase in the production and availability of single-cell data. The ability integrate a set multi-technology measurements would allow identification biologically or clinically meaningful observations through unification perspectives afforded by each technology. In most cases, however, profiling technologies consume used cells thus pairwise correspondences between datasets are lost. Due sheer size can acquire, scalable algorithms that able...
Pancreatic adenocarcinoma (PDAC) epitomizes a deadly cancer driven by abnormal KRAS signaling. Here, we show that the eIF4A RNA helicase is required for translation of key signaling molecules and pharmacological inhibition has single-agent activity against murine human PDAC models at safe dose levels. EIF4A was uniquely mRNAs with long highly structured 5' untranslated regions, including those multiple G-quadruplex elements. Computational analyses identified these features in encoding...
CD8+ tumor-infiltrating T cells can be regarded as one of the most relevant predictive biomarkers in immune-oncology. Highly infiltrated tumors, referred to inflamed (clinically "hot"), show favorable response immune checkpoint inhibitors contrast tumors with a scarce infiltrate called desert or excluded "cold"). Nevertheless, quantitative and reproducible methods examining their prevalence within are lacking. We therefore established computational diagnostic algorithm quantitatively measure...
ABSTRACT Motivation Deep learning techniques have yielded tremendous progress in the field of computational biology over last decade, however many these are opaque to user. To provide interpretable results, methods incorporated biological priors directly into task; one such prior is pathway structure. While pathways represent most processes cell, high level correlation and hierarchical structure make it complicated determine an appropriate representation. Results Here, we present module...
Abstract We present the most comprehensive catalogue of cancer-associated gene alterations through characterization tumor transcriptomes from 1,188 donors Pan-Cancer Analysis Whole Genomes project. Using matched whole-genome sequencing data, we attributed RNA to germline and somatic DNA alterations, revealing likely genetic mechanisms. identified 444 associations expression with non-coding single-nucleotide variants. found 1,872 splicing associated mutation in intronic regions, including...
Abstract The ability to understand and predict molecular responses towards external perturbations is a core question in biology. Technological advancements the recent past have enabled generation of high-resolution single-cell data, making it possible profile individual cells under different experimentally controlled perturbations. However, are typically destroyed during measurement, resulting unpaired distributions over either perturbed or non-perturbed cells. Leveraging theory optimal...
Abstract Motivation Several recently developed single-cell DNA sequencing technologies enable whole-genome of thousands cells. However, the ultra-low coverage sequenced data (<0.05× per cell) mostly limits their usage to identification copy number alterations in multi-megabase segments. Many tumors are not number-driven, and thus single-nucleotide variant (SNV)-based subclone detection may contribute a more comprehensive view on intra-tumor heterogeneity. Due low data, SNVs is only...
Abstract Understanding and predicting molecular responses in single cells upon chemical, genetic, or mechanical perturbations is a core question biology. Obtaining single-cell measurements typically requires the to be destroyed. This makes learning heterogeneous perturbation challenging as we only observe unpaired distributions of perturbed nonperturbed cells. Here leverage theory optimal transport recent advent convex neural architectures present CellOT, framework for response individual...
The recent adoption of Electronic Health Records (EHRs) by health care providers has introduced an important source data that provides detailed and highly specific insights into patient phenotypes over large cohorts. These datasets, in combination with machine learning statistical approaches, generate new opportunities for research clinical care. However, many methods require the representations to be structured formats, while information EHR is often locked unstructured texts designed human...
A bstract Motivation Recent technological advances have led to an increase in the production and availability of single-cell data. The ability integrate a set multi-technology measurements would allow identification biologically or clinically meaningful observations through unification perspectives afforded by each technology. In most cases, however, profiling technologies consume used cells thus pairwise correspondences between datasets are lost. Due sheer size can acquire, scalable...
Multimodal profiling strategies promise to produce more informative insights into biomedical cohorts via the integration of information each modality contributes. To perform this integration, however, development novel analytical is needed. often come at expense lower sample numbers, which can challenge methods uncover shared signals across a cohort. Thus, factor analysis approaches are commonly used for high-dimensional data in molecular biology, they typically do not yield representations...
Abstract New and effective therapeutics are urgently needed for the treatment of pancreatic ductal adenocarcinoma (PDAC). The eIF4A/DDX2 RNA helicase drives translation mRNAs with highly structured 5′UTRs. natural compound silvestrol synthetic analogues potent selective inhibitors eIF4A1/2 that show promising activity in models hematologic malignancies. Here, we have nanomolar against PDAC cell lines organoids vitro. Moreover, see single-agent KRAS/p53 mouse model also xenograft primary,...
Abstract Pancreatic cancer is one of the most aggressive cancers with no targeted therapy available. RNA translation activated in pancreatic and at same time also refractory to mTor inhibition. We have used a inhibitor for eIF4A, helicase that downstream mTOR signaling can be functionally cancer. establish Silvestrol its analog CR-31B showed potent anti-tumor activity cell lines vitro vivo. Silvestrol/CR-31B reduced cells organoids growth derived from mouse model human patient samples vitro....
We present a novel probabilistic clustering model for objects that are represented via pairwise distances and observed at different time points. The proposed method utilizes the information given by adjacent points to find underlying cluster structure obtain smooth evolution. This approach allows number of clusters differ every point, no identification on identities is needed. Further, does not require being specified in advance -- they instead determined automatically using Dirichlet...
Pediatric high grade gliomas (pHGG) are among the most common malignant brain tumors in childhood and account for majority of cancer related mortality this age group. The recent discovery two recurrent mutations tail histone variant H3.3, resulting either a substitution Glycine at position 34 (G34R/V) or Lysine 27 (K27 M), has deepened our understanding tumor entity led to new molecular classification system. However, functional consequences these on biology remain largely unknown....