Joshua M. Stuart

Transformer-based modeling of Clonal Selection and Expression Dynamics (TraCSED) reveals resistance signatures in breast cancer

TraCSED identifies resistance pathways in breast cancer by analyzing scRNA-seq data and clonal selection dynamics over time.

A Spatiotemporal and Machine-Learning Platform Accelerates the Manufacturing of hPSC-derived Esophageal Mucosa

The paper develops a spatiotemporal multi-omics platform and machine-learning tool to manufacture hPSC-derived esophageal mucosa.

Associating Transcription Factors to Single-Cell Trajectories with DREAMIT

DREAMIT identifies transcription factor activity patterns along single-cell trajectories, showing improved tissue-specific performance over existing methods.

Dual RNA-Seq analysis of SARS-CoV-2 correlates specific human transcriptional response pathways directly to viral expression

This study uses dual RNA-seq to analyze co-regulation of human and SARS-CoV-2 genes, identifying tissue-specific expression patterns and potential therapeutic targets.

Towards Inferring Nanopore Sequencing Ionic Currents from Nucleotide Chemical Structures

The paper presents a deep learning framework to predict nanopore sequencing ionic currents from nucleotide chemical structures and demonstrates its ability to generalize chemical information for modified nucleotides.

Patient Informed Domain Adaptation Improves Clinical Drug Response Prediction

This paper develops a deep learning framework to improve clinical drug response prediction by integrating gene expression and drug chemical structures, showing better performance than existing methods.

Highly Accurate Cancer Phenotype Prediction with AKLIMATE, a Stacked Kernel Learner Integrating Multimodal Genomic Data and Pathway Knowledge

AKLIMATE predicts cancer phenotypes by integrating multi-omic data and pathway knowledge, outperforming existing methods in tasks like microsatellite instability prediction and survival analysis.

Biological network-inspired interpretable variational autoencoder

This paper introduces VEGA, a variational autoencoder with interpretable latent variables inspired by biological networks, demonstrating its ability to model cellular responses and regulatory mechanisms.

Modeling human TBX5 haploinsufficiency predicts regulatory networks for congenital heart disease

The paper models TBX5 haploinsufficiency in iPSC-derived cardiomyocytes to identify gene regulatory network imbalances linked to congenital heart disease.

Creating Standards for Evaluating Tumour Subclonal Reconstruction

This study develops quantitative metrics and simulated datasets to evaluate subclonal reconstruction methods in tumours.

Using Transcriptional Signatures to Find Cancer Drivers with LURE

The paper introduces LURE, a semi-supervised method to predict cancer drivers using gene expression signatures, and applies it to TCGA data to find new associations with known mutation events.

Biological Process Activity Transformation of Single Cell Gene Expression for Cross-Species Alignment

This paper transforms single-cell RNA-Seq data into biological process activities to enable cross-species alignment of cellular states.

Combining accurate tumour genome simulation with crowd-sourcing to benchmark somatic structural variant detection

The paper benchmarks somatic structural variant detection methods using synthetic tumors and crowd-sourced submissions, finding that ensembles do not always outperform individual methods.

Valection: Design Optimization for Validation and Verification Studies

Valection optimizes the selection of verification candidates to estimate error profiles in genomics studies by evaluating different strategies on synthetic data.

Pathway and network analysis of more than 2,500 whole cancer genomes

The paper analyzes non-coding mutations across 2,583 cancer genomes to identify genes and pathways affected, finding clusters of interacting proteins and insights into cancer mechanisms.

Germline Contamination and Leakage in Whole Genome Somatic Single Nucleotide Variant Detection

The paper quantifies germline leakage in somatic SNV detection across 259 datasets and introduces a tool to correct it, finding low leakage rates that can be further reduced with pre-filtering.

Discovery and characterization of coding and non-coding driver mutations in more than 2,500 whole cancer genomes

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

Tracing co-regulatory network dynamics in noisy, single-cell transcriptome trajectories

SCIMITAR infers progression trajectories and regulatory networks from single-cell transcriptomes, identifying progression-associated genes and co-regulatory states in neural differentiation.