Stephen Piccolo

This paper examines how batch effects impact machine learning classifier performance across multiple studies and compares adjustment methods.

This paper evaluates using language models to improve gene-expression dataset discovery in GEO by comparing embeddings to GEO's search engine.

This study performs a meta-analysis of breast cancer gene expression across diverse populations, identifying novel genes associated with race, hormone receptor status, and tumor subtypes.

The paper evaluates color accessibility of scientific images for deuteranopes, finds 12.8% are problematic, and develops a tool to identify and simulate these issues.

TidyGEO tidies and reformats GEO data for analysis by cleaning sample annotations and integrating them with assay data.

The paper curates 102 breast cancer gene-expression datasets and provides a computational pipeline for data processing and metadata annotation to facilitate reuse and comparative analysis.

The paper analyzes DNA methylation patterns in five pediatric tumor types, finding hypermethylation is more common than hypomethylation and identifying gene methylation patterns associated with patient subsets.

The paper benchmarks 50 classification algorithms on 50 gene-expression datasets to evaluate their performance in patient subgroup classification.

The paper introduces ToolJig, a web application that simplifies creating and validating Common Workflow Language (CWL) documents for reproducible bioinformatics workflows.

This study predicts cancer cell drug sensitivity using DNA methylation data from cell lines and evaluates model performance on both cell lines and human tumors.

ShinyLearner is a tool for benchmarking machine-learning classification algorithms on tabular data, enabling transparent and reproducible comparisons across software packages.

The paper benchmarks methods for querying large tabular datasets and proposes a coordinate-based, fixed-width storage format for efficient data extraction.

The paper groups genomic aberrations by their transcriptional effects to identify drug-repurposing candidates and known treatment-predictive gene pairs.

This paper presents a crowdsourced challenge to predict early discontinuation of docetaxel in mCRPC patients using clinical data from multiple trials, finding that ensemble models outperform individual models.