Priya Ranjan from Oak Ridge National Laboratory was the next speaker I watched as part of the KBase Microbial Community Modeling Workshop recordings. The title of this session was “Pairwise analysis tools between strains” that would be very useful for us! Rnajan is collaborating with the Plant-Microbe Interfaces (PMI) project and is designing a series of apps. As part of PMI, over five hundred isolates have been sequenced and co-culture experiments have been conducted. Ranjan and team are leveraging machine learning and metabolic modeling to predict pairwise interactions and phenotypes. The apps that have been developed as part of this project are Calculate Metabolic Interaction Scores, Annotate BGCs, and Predict Antimicrobial Activity. CommScores is the first app and calculates community scores. The fraction of metabolic resource overlap (MRO, competition) is calculated. Next, metabolic interaction potential (MIP) collaboration is determined. Finally, growth yield difference (GYD) is calculated. The second app is antiSMASH V7 that provides antiSMASH in KBase. The app identifies biosynthetic gene clusters (BGC) and uses a machine-learning workflow to predict antimicrobial activity from a microbial genome. Ranjan demonstrated the apps. The CommScores results include a heat map to compare scores across different pairs. The second app runs AntiSMASH V7 and produces a tabular view with each row for a genome and columns for different categories of biosynthetic gene clusters. The app also generates a rich html report with the gene clusters identified depicted. The BCG app generates a table for the potential of antimicrobial activity. Rajan noted that they will now take experimental data for validation and improvement of the apps. The team is also building the machine learning framework in KBase. I am interested in running several of these tools with new genomes!
