Phenotyping at Scale: Turning Microbial Growth into AI-Ready Data

Phenotyping at Scale: Turning Microbial Growth into AI-Ready Data at Society of Biological Engineering's 8th Commercializing Industrial Biotechnology (CIB) in Chicago, IL USA

Abstract: ExFAB, the NSF-funded BioFoundry for Extreme and Exceptional Microbes, is building the experimental–computational pipelines required for AI- and ML-enabled biology — from automated data generation to predictive models. A central focus is creating high-quality, large-scale genotype–phenotype datasets using standardized automated workflows and pooled screens. ExFAB’s efforts in this space began with creating ML/AI suitable datasets for species-specific CRISPR guide design models. Key lessons learned from this work were that high quality, well-annotated, and balanced datasets are needed for accurate modeling and predictions. These lessons learned have informed our genotype-phenotype pipelines. ExFAB’s automated workflows are capable of characterizing and testing large mutants and isolate libraries in high throughput. This is accomplished with robotic arraying and automated time-course imaging, ultimately producing tens of thousands of images that capture microbial colonies as they grow under various conditions. The libraries are sufficiently sized to ensure the capture of well-balanced datasets, a laboratory information management system (LIMS) ensures the capture of critical metadata, and a new image analysis platform — PhenoTypic — extracts quantitative growth, color, and morphology data. This talk first presents our lessons learned in generating AI-ready data, then presents a series of case studies in the use of PhenoTypic and ExFAB’s high throughput phenotyping workflows, including the study of thermotolerant Kluyveromyces yeasts, metabolite production in environmental isolates of the basidiomycetes fungi Rhodoturula, and morphological study of wood-decaying Ganoderma fungi.