Haitham Elmarakeby

The paper expands P-NET, a deep learning framework that uses biological pathways for genomic and transcriptomic predictions, showing competitive performance and interpretability.

The paper evaluates how dataset characteristics affect biologically-informed neural network performance using simulations and empirical testing with prostate cancer data.

This study evaluates twelve single-cell foundation models for cancer-related tasks and finds they offer limited clinical outcome prediction advantages over simpler models.

This paper introduces a biologically informed deep learning model for prostate cancer classification and identifies molecular drivers of treatment resistance.

This paper introduces a systematic auditing framework to identify biases in machine learning applications in biology, showing that biases can hinder model generalizability and lead to poor performance when training data lacks sufficient signal.

The paper presents a large-scale cancer pharmacogenomic screen with 11,576 experiments on 910 drug combinations across 85 cancer cell lines, enabling crowd-sourced prediction of synergistic drug pairs and biomarkers.