Session: Session 2

Towards a Unified Pipeline: Comparing tools for Thermal Proteome Profiling experiments

Karen DRUART¹, Eléonore BOUSCASSE¹, Rayen ELJ¹, Angélique AMO², Paola B. ARIMONDO³, Jonathan B. WEITZMAN², Mariette MATONDO¹

¹Institut Pasteur, Proteomics Core Facility, MSBio UtechS, UAR CNRS 2024, Université Paris Cité, Paris, France
²Epigenetics and Cell Fate, CNRS UMR7126, Université Paris Cité, Paris, France
³Institut Pasteur, Epigenetic Chemical Biology, CNRS UMR3523 Chem4Life, Université Paris Cité, Paris, France

Epigenetic modifications, targeting DNA or histone proteins, modulate gene expression without altering the DNA sequence, and are potentially reversible. This makes epigenetic modifying enzymes promising drug targets against antimicrobial resistance (AMR). Our research suggests that targeting epigenetic changes offers an innovative strategy to counteract pathogen resistance.

Characterizing the targets of epigenetic drugs is challenging. One promising approach is Thermal Proteome Profiling (TPP), derived from the Cellular Thermal Shift assay (CETSA), which identifies drug-affected proteins across the proteome. Over the past decade, various methods for analysing TPP datasets have been developed, resulting in issues such as overfitting and a lack of consensus on best practices. To address this, we applied several analytical methods to a well-known set of drug-target pairs to compare the proteins identified.

Our study builds on the seminal TPP analysis by the Savitski lab (Franken et al., 2015) using K562 human cells treated with the lysine deacetylase inhibitor Panobinostat. We meticulously replicated their experimental design and analysed the raw data with MaxQuant, then re-analysed the published raw files to compare results with Mascot and evaluate the impact of the search engine. Our bioinformatics workflow utilized the flexible TPP package by Childs et al. (2004), initially developed with the Panobinostat dataset.

We compared the proteins identified across three different analytical pipelines to assess the impact of the search engine on protein identification. We also tested various TPP package parameters, such as normalization and thermal-shift estimation techniques. Our results indicate that the choice of search engine significantly affects the proteins identified and highlights low reproducibility with the TPP package. This underscores the importance of selecting appropriate analytical methods to ensure reliable identification of drug targets in proteomic studies.