Advanced Computational Core

CORE 1: ADVANCED COMPUTATIONAL CORE SUMMARY / ABSTRACT The Gladstone PhAIge Therapy Center tackles the challenge of rationally designing phage therapies by developing high-throughput assays for quantifying each step of the phage infection cycle (Project 1) and precisely matching phage gene repertoires to bacterial strains (Project 2). We focus on multi-drug resistant Klebsiella pneumoniae (Kp), a leading cause of pneumonia and other hospital-acquired pulmonary diseases, and leverage human lung organoids (Core 2) to capture interactions that are missed in traditional assays. The resulting sequencing-based platform technologies will generate unprecedented information about the efficacy of Kp phage and will be easily extendable to other pathogens in the ESKAPE group and beyond. To accelerate the development of predictive assays in the Gladstone PhAIge Therapy Center, we propose an Advanced Computational Core (Core 1) that will provide cutting-edge expertise in AI, bioinformatics, and mathematical modeling. The core's overall objective is to take the guess work out of phage assays by developing predictive understanding of phage-bacteria interactions. Our trained models and open-source code will be freely shared as key components of the assays generated in the Gladstone PhAIge Therapy Center. The key innovation of the Advanced Computational Core is our utilization of AI modeling and GPU computing to design and interpret assay experiments. Leveraging recent advances in protein structure and function prediction using large language models, we will propose sequence variants for testing in both Projects (Aim 1). Our AI models will recommend both naturally occurring and synthetic sequence variants, with a range of predicted activity levels, that will make assay development more efficient than current trial-and-error approaches. Furthermore, iteratively predicting, testing, and fine-tuning AI models in collaboration with both Projects will enable us to develop a mechanistic understanding of how phage-bacterial interactions depend on different genomic features (Aim 2). As a result, we will be able to make accurate predictions about the efficacy of newly discovered phage against specific bacterial strains based on genomic features, such as accessory genes (AGs) (Project 2) and receptor binding protein (RBP) variants (Project 1). Because the Center's assays utilize advanced sequencing technologies, including long reads and barcoded libraries for pooled screeding, the Core's expert bioinformatics support will be essential for ensuring robust quantification of phage success across assay platforms. This will include computing estimates of rate constants for each individual stage of infection, from adsorption to lysis, along with resampling-based quantification of uncertainty in these estimates (Aim 3). Together, these strategies will accelerate assay development, enable rational selection of natural phage for each patient based on genomic sequencing, and ultimately provide the design rules for engineering therapeutic phage.