Searcharxiv⌕ Search

arXiv subjects

Josh V. Vermaas

Publications and source records attributed to Josh V. Vermaas.

4 recordsLinked to original sources

MathAgent: Multi-Agent Optimization of Mathematical Invariants for Molecular Property Prediction

Mathematical invariants provide complementary representations of molecular structure, but their predictive utility can vary across datasets and prediction tasks. For a new dataset, appropriate construction of invariant representations from algebraic topology, differential geometry, topological spectral theory, and commutative algebra is a challenging task, particularly when evaluating multiple representations is computationally expensive. We develop a Mathematical Agentic Model (MathAgent) to automatically optimize invariant representations in molecular property prediction. The framework combines dataset-aware representation design, low-cost candidate screening, and progressive evaluation. Language-model reasoning supports task interpretation, strategy proposal, and workflow coordination, while scientific computations and numerical decisions are carried out by deterministic tools. We evaluate MathAgent on protein--ligand binding-affinity and quantitative toxicity prediction tasks. Different datasets favor different combinations of mathematical invariants, demonstrating the need for dataset-adaptive representation selection rather than a fixed universal representation. Objective-conditioned executions further show that the framework can adjust its evaluation strategy according to user-specified goals. MathAgent therefore provides a systematic, evidence-driven, and traceable approach for optimizing mathematical invariant representations and reducing reliance on manual configuration and/or experimentation in molecular property prediction.

q-bio.BM↗

AlphaFunctor: Bridging The Gap Between Protein Function Annotation and Property Prediction

The fundamental relationship among protein sequence, structure, function, and physicochemical properties is a central principle in biology. While in principle protein function and properties should be able to be derived directly from protein sequence, in practice protein function and property prediction methods have been designed around specific datasets and specific property or function subsets, leading to an enormous gap between function annotation and property prediction. To address these challenges, we introduce AlphaFunctor, a category theory based foundation model-like platform to bridge the gap between protein function annotation and property prediction. Based on the hypothesis that protein function and properties can be directly derived from protein sequence, AlphaFunctor predicts protein functions as represented by Gene Ontology terms directly from sequence. Using these function predictions, AlphaFunctor further maps protein functions using topological spectral theory, path-complex neural networks, and protein domain analysis onto downstream property prediction. AlphaFunctor is (pre)trained in nearly 0.6 million protein function data points to deliver the state-of-the-art protein function annotation on three benchmark datasets. Without task-specific network redesign, AlphaFunctor maps qualitative protein function annotation to various qualitative and quantitative protein property predictions, outperforming other dataset-specific and task-specific competing predictors.

q-bio.BM↗

GPU-Accelerated Drug Discovery with Docking on the Summit Supercomputer: Porting, Optimization, and Application to COVID-19 Research

Protein-ligand docking is an in silico tool used to screen potential drug compounds for their ability to bind to a given protein receptor within a drug-discovery campaign. Experimental drug screening is expensive and time consuming, and it is desirable to carry out large scale docking calculations in a high-throughput manner to narrow the experimental search space. Few of the existing computational docking tools were designed with high performance computing in mind. Therefore, optimizations to maximize use of high-performance computational resources available at leadership-class computing facilities enables these facilities to be leveraged for drug discovery. Here we present the porting, optimization, and validation of the AutoDock-GPU program for the Summit supercomputer, and its application to initial compound screening efforts to target proteins of the SARS-CoV-2 virus responsible for the current COVID-19 pandemic.

q-bio.BM↗

Beyond the Boltzmann factor for corrections to scaling in ferromagnetic materials and critical fluids

The Boltzmann factor comes from the linear change in entropy of an infinite heat bath during a local fluctuation; small systems have significant nonlinear terms. We present theoretical arguments, experimental data, and Monte-Carlo simulations indicating that nonlinear terms may also occur when a particle interacts directly with a finite number of neighboring particles, forming a local region that fluctuates independent of the infinite bath. A possible mechanism comes from the net force necessary to change the state of a particle while conserving local momentum. These finite-sized local regions yield nonlinear fluctuation constraints, beyond the Boltzmann factor. One such fluctuation constraint applied to simulations of the Ising model lowers the energy, makes the entropy extensive, and greatly improves agreement with the corrections to scaling measured in ferromagnetic materials and critical fluids.

cond-mat.stat-mech↗