Lincoln David Stein

Apollo 3 is a new manual genome annotation tool that integrates with JBrowse 2, enabling multi-genome editing and synteny visualization.

This study identifies metabolic partnerships between medulloblastoma and macrophages that drive leptomeningeal metastases and suggests targeting the CXCL12-SCARB1-cholesterol axis as a therapeutic strategy.

The paper identifies ependymomas as originating from the pre-neural crest/roof plate lineage using single-cell transcriptomics and lineage tracing.

This paper applies large language models to assist in gene annotation into Reactome pathways, showing promising results in supporting curation workflows.

React-to-Me, a retrieval-augmented chatbot grounding GPT-4o-mini in curated Reactome pathways with hyperlinked citations, earned higher blinded expert ratings than ungrounded GPT-4o-mini (odds ratio 2.01).

The paper introduces OncoGAN, a generative AI tool that creates synthetic cancer genomes to address the lack of gold-standard datasets for benchmarking cancer genome analysis tools.

OmniPert predicts responses to genetic and chemical perturbations in single cancer cells using a deep learning model trained on diverse cellular profiles.

This paper introduces a framework to interpret attention mechanisms in genomic transformer models by correlating attention scores with biological annotations, showing how biological features are learned through pre-training and fine-tuning.

The paper shows BRCA1 correlates with proliferation in human neural progenitor cells and medulloblastoma, particularly in Group 3 and 4 tumors, and is linked to worse prognosis.

This paper identifies how PVT1 rearrangements in MYC+ cancers generate oncogenic peptides that synergistically activate MYC, AKT, and RAS-MAPK pathways, enhancing tumor aggressiveness.

The paper introduces FHR, a metadata standard for reference genome data to improve interoperability and data provenance tracking across platforms.

The paper annotates disease variants in the context of biological pathways to infer molecular phenotypes and model variant impacts.

This paper introduces a multi-modal deep learning algorithm for drug response prediction that improves generalizability by integrating diverse cell line data and addressing data skewness.

The paper updates the Alliance of Genome Resources' central infrastructure, adding tools and enhancing data management and curation capabilities.

This paper explores using ChatGPT in Reactome curation to automate pathway annotation, comparing its outputs to manual curation and identifying challenges and improvements.

The paper predicts functional interactions of dark proteins with Reactome pathways using a random forest classifier and develops a web portal to visualize these interactions.

JBrowse 2 is a genome browser with enhanced views for structural variation and synteny, offering modular design and multiple usage options.

JBrowse Jupyter provides a Python interface to JBrowse 2 for interactive genomic visualization within Jupyter notebooks.

Pairtree reconstructs clone trees from multiple bulk DNA samples to infer cancer evolutionary histories with higher accuracy and scalability.

JBrowse Connect enables live messaging, notifications, and user-initiated analyses in JBrowse via WebSockets, with JBlast as an example application for BLAST searches integrated with Galaxy.

The paper quantifies immune-based counterselection of somatic mutations by analyzing their depletion in MHC class I-displayed peptides and shows that MHC alleles are incompletely dominant.

The paper predicts cancer cell-of-origin from DNA sequences by matching mutational landscapes to chromatin modification patterns in normal tissues.

The paper compares whole-genome data from primary pancreatic cancer, xenografts, and organoids to assess genetic concordance across different models.

The paper identifies and characterizes coding and non-coding driver mutations across 2,500 cancer genomes, revealing new candidates and questioning previous findings.

DriverPower identifies cancer driver mutations by combining mutational burden and functional impact scores, outperforming existing methods in coding and non-coding driver discovery.

The paper describes the PCAWG project's large-scale analysis of cancer whole genomes across multiple computing environments, producing validated consensus variants for downstream research.