Rapid processing of high-throughput thermal shift assay data for lead compound identification

Authors

• SC Waterworth
• SS Kurian
• ND Sharma
• C Wolcott
• DE Donohue
• BR O'Keefe
• JA Beutler

Abstract

Thermal Shift Assays (TSA), also known as Differential Scanning Fluorimetry (DSF), have been used to determine the melting temperature of a protein. Addition of a ligand may influence the melting temperature of a protein by either stabilizing or destabilizing the native structure, resulting in the characteristic “shift” of the melting temperature. TSA can therefore be used as an initial exploratory assay to detect potentially novel protein-compound interactions. In the context of a 384-well plate setup, each plate generates around 100,000 data points. A single run can involve hundreds of plates, resulting in tens of millions of data points. Analyzing this substantial volume of data becomes even more challenging when using natural product extracts, as the data tends to be not only noisier but often deviates from the expected sigmoidal distribution. To address these challenges, we developed an algorithm that can handle this complex data. Briefly, the data for each well is “cleaned” by extracting only the sigmoidal region(s). Each sigmoidal region is then fitted against a Boltzmann sigmoid model and the melting temperature(s) are calculated. Experimental wells with large thermal shifts, relative to the plate controls are candidates for further testing. When using only 4 CPUs on a standard MacBook Pro, it takes 27 minutes to process 56 plates (~5.6 million data points, 19,712 compounds tested). A companion visualization tool has been made available for additional user QC and effortless hit identification.

Scientific Focus Area: Computational Biology

This page was last updated on Tuesday, August 6, 2024