(Click titles to navigate to project homepage)
My research mainly focuses on translational bioinformatics, with methodological contributions in artificial intelligence (especially graph machine learning and automated machine learning), and applications in computational toxicology and systems pharmacology. This page lists some of the major research projects I'm involved in.
A toolchain for artificial intelligence applications in computational toxicology. ComptoxAI provides a number of interconnected components, including a large heterogeneous graph database of toxicology concepts, an OWL ontology for computational toxicology, a gallery of graph machine learning models, and informational/educational resources.
An extension to TPOT (a Python tool for optimizing machine learning pipelines using genetic programming) that adds support for PyTorch neural network models. We're currently developing support for diverse data types, including imaging data and sequence data (free text, time series, etc.). Our goal is to introduce neural networks in the context of automated machine learning without losing the model introspection advantages that come with AutoML run on non-neural network algorithms.
VenomSeq is a sequencing and data analysis pipeline for discovering therapeutic effects from venoms and venom peptides. Briefly, VenomSeq involves exposing human cell cultures to dilute concentrations of venoms, then performing RNA-Seq to determine which genes are up- and down-regulated in response to the venom. These expression signatures are then compared to expression signatures produced by drugs with known effects (taken from the LINCS L1000 / Connectivity Map resource), and used to propose therapeutic hypotheses for the venoms based on observed similarities. These hypotheses can then be evaluated experimentally to determine whether the therapeutic effect is plausible.
VenomSeq has currently been performed on 25 diverse crude venoms and 11 purified venom proteins from the venomous auger snail. We are nearing journal submission of VenomSeq's first major publication and will update here accordingly!
A knowledge base and collection of related tools for computational venom research, particularly for the purpose of discovering new therapeutic effects of venom-derived peptides.
An OWL ontology used to structure and perform automated reasoning on VenomKB, enabling advanced computational queries of venom data and discovery of previously unknown therapeutic associations between venoms and human diseases.