meiler lab / vanderbilt computational structural biology
building the computational backbone
for a $46m vaccine project
may 2025 - present / remote from seattle / computational pipeline development
The Meiler Lab received a $46M ARPA-H grant to stabilize alphaviruses for vaccine development. I'm building the workflow that helps the computational team generate their protein structures as efficiently as possible: benchmarking prediction methods, automating relaxation protocols, and streamlining sequence design so they can iterate faster.
methods / pipeline
Prediction, relaxation, sequence design, and validation in one automated workflow.
structure prediction
Ran AlphaFold 2.3.2 and Boltz-1 v0.4.1 across the full BM5.5 protein-protein docking benchmark (257 complexes). Generated initial structure predictions for all bound and unbound conformations.
AMBER relaxation
Applied AMBER force field relaxation on all AlphaFold and Boltz-1 predicted models. Physics-based energy minimization resolves steric clashes and refines local geometry.
Rosetta relaxation
Ran 6 Rosetta 3.15 relaxation protocols with 5 replicates each across all 257 complexes. Systematic approach tests multiple energy landscapes and sampling strategies.
quality validation
Validated all structures using MolProbity quality metrics (clashscore, Ramachandran outliers, rotamer analysis) and PoseBusters geometry validation (bond lengths, angles, steric clashes).
results
Benchmarking outcomes and pipeline validation results.
full BM5.5 benchmark
First systematic comparison of AI-predicted vs. experimentally-derived structures at this scale with multi-protocol Rosetta relaxation across all 257 complexes.
257 complexes / 6,800+ structuresmulti-protocol relaxation
Systematic evaluation of 6 distinct Rosetta relaxation protocols, each with 5 independent replicates. Enables robust statistical comparison of protocol performance across diverse complex types.
6 protocols x 5 replicates x 257 complexesstructure quality metrics
MolProbity validation across all structures: clashscore, Ramachandran analysis, rotamer quality, and Cbeta deviations. PoseBusters geometry checks confirm physical plausibility post-relaxation.
molprobity + posebusters on all modelsprotein design pipeline
End-to-end integration of structure prediction, relaxation, and inverse folding for alphavirus vaccine stabilization. Automated from target complex through validated designed sequences.
alphafold + rosetta + proteinmpnnpublications
BM5.5 protein complex benchmarking
First-author. AI structure prediction vs. experimental benchmarking with multi-protocol relaxation. Full publication to follow.
alphavirus stabilization for vaccine design
Contributing author. Computational pipeline for the ARPA-H vaccine project. Meiler Lab, Vanderbilt.
Protein_Data_Analysis
Validation and analysis pipeline for protein structure quality metrics. MolProbity scoring, PoseBusters geometry checks, and comparative statistics.
Protein_Relax_Pipeline
Rosetta and AMBER relaxation framework. Automated pipeline for running multiple protocols with configurable replicates. Built for HPC batch submission.
Protein_Ideal
Full BM5.5 benchmark dataset. 257 complexes across AlphaFold 2.3.2 and Boltz-1 v0.4.1. Replicates the Protein_Relax_Pipeline with relaxation results and analysis.