Computational biology researcher working on AI, protein design, and structure-based drug discovery.
UWaterloo Biotech/Econ '26. Previously SickKids, OICR, and QDX Technologies.
9 candidates from 5,793 proteins screened for surface exposure, nucleotide diversity, and ranked by abundance; binder design next
Topics
Tools
0.91 AUC predicting PPI disruption, from 47,000+ AlphaFold-Multimer jobs
Topics
Tools
74% balanced accuracy across 14 CATH architectures on held-out superfamilies, level with Foldseek
Topics
Tools
Read the write-up: PAC-MAPS: a protein architecture classifier CNN that matches Foldseek on unseen protein folds Code for PAC-MAPS: a protein architecture classifier CNN that matches Foldseek on unseen protein folds on GitHub
71 candidates pulled from the COCONUT library, docked with GNINA and ranked on CNN-VS
Topics
Tools
Read the write-up: 71 natural alternatives to a $200M diabetes drug found entirely in silico using GNINA Code for 71 natural alternatives to a $200M diabetes drug found entirely in silico using GNINA on GitHub
79.3% sensitivity on peptide binding sites, 12 points above the original paper's model, at 78.6% accuracy
Topics
Tools
Read the write-up: Visual-PepBDB: a CNN that beats the published model's sensitivity by 12 points Code for Visual-PepBDB: a CNN that beats the published model's sensitivity by 12 points on GitHub