ProTargetMiner: A proteome signature library of anticancer molecules for functional discovery

Abstract

We present a publicly available, expandable proteome signature library of anticancer molecules in A549 adenocarcinoma cells. Based on 287 proteomes affected by 56 drugs, the main dataset contains 7,328 proteins and 1,307,859 refined protein-drug pairs. By employing the specificity concept in partial least square modeling, deconvolution of drug targets and mechanistic proteins is achieved for most compounds, including some kinase inhibitors. We built the first protein co-regulation database that takes into account both protein expression and degradation. A surprising number of strong anti-correlations is found, underscoring the importance of protein repression in cell regulation. Our analysis uncovered a group of proteins with extremely steady expression which are likely essential for core cellular functions. These findings bring about deeper understanding of cell mechanics. Extension of the dataset to novel compounds will facilitate drug design. The introduced specificity concept and modeling scheme are beneficial in other analysis types as well.Statement of SignificanceProTargetMiner is the first of its kind library of proteome responses of human cancer cells to anticancer molecules. This expandable resource facilitates the deconvolution of drug targets, action mechanisms, and cellular effects. It reveals death modalities, uncovers protein co-regulation and anti-correlation networks and defines the “untouchable” proteome essential for core cellular functionalities.

The deubiquitinase inhibitor b-AP15 induces strong proteotoxic stress and mitochondrial damage

“…Therefore, even proteins with minimal but reproducible changes can be identified using this strategy. Figure reused with permission from (Saei, Chernobrovkin et al 2018).…”

“…LC50 values were determined for the compound library (n=118) plus some other compounds in three cell lines; A549 cells were chosen based on their higher sensitivity; compounds were selected based on their cytotoxicity; cells were treated with 56 compounds and samples were prepared for shotgun proteomics in at least three independent biological replicates; samples were multiplexed in each experiment (methotrexate, paclitaxel and camptothecin were included in all 9 experiments); samples were lysed, digested, and labeled with TMT-10plex; pooled within each experiment and fractionated to increase the proteome coverage; individual fractions were analyzed by LC-MS/MS, followed by protein identification, quantification and data post-processing. Figure reused with permission from (Saei, Chernobrovkin et al 2018).…”

“…We would like to envision that these pathways are represented by cell death trajectories in the proteome space, which encompasses all possible cell states. We define cell death trajectory as a track in the proteome space, which is caused by a cytotoxic agent and passes from a normal living state to a death state (Saei, Chernobrovkin et al 2018). The proteome space and death trajectories are shown in schematics in Figure 16.…”

“…Disconnected proteins not involved in the pathway has been omitted. Figure adapted with permission from (Saei, Chernobrovkin et al 2018).…”

“…NDUFV2 and CARS2 proteins were randomly selected as non-target proteins. Figure reused with permission from (Saei, Chernobrovkin et al 2018).…”