Reconstructing targetable pathways in lung cancer by integrating diverse omics data

We profile transcriptome, proteome and phosphoproteome in a panel of non-small cell lung cancer (NSCLC) cell lines in order to reconstruct targetable networks associated with KRAS dependency. We develop a two-step bioinformatics strategy addressing the challenge of integrating these disparate datasets. We first define an “abundance-score” combining transcript, protein and phospho-protein abundances to nominate differentially abundant proteins and then use the Prize Collecting Steiner Tree algorithm to identify functional sub-networks. We identify three modules centered on KRAS and MET, LCK and PAK1 and Beta-catenin. We validate activation of these proteins in KRAS-dependent cells and perform functional studies defining LCK as a critical gene for cell proliferation in KRAS-dependent but not KRAS-independent NSCLCs. These results are the first evidence that suggest LCK as a potential druggable target protein in KRAS-dependent lung cancers. Data analysis: Raw spectra files were converted to mzXML using ReadAW. The mzXML files were searched using X!Tandem with the k-score plug-in. The proteomic searches were performed using the following options: allow up to 2 missed tryptic cleavages, a parent ion tolerance window of -1 to +4 Daltons, and a fragment ion tolerance of 0.8 Da. The following variable modifications were allowed: phosphorylation of Serine, Threonine, and Tyrosine (+79.966331@[STY]), oxidation of Methionine (+15.994920@M), and carbamidomethylation of Cysteine (+57.021460@C). All protein searches were performed using the Human Refseq protein database (release 47). Appended to this database were common proteomic contaminants and reversed protein sequences to serve as decoys. The X!Tandem results were then post processed with PeptideProphet and ProteinProphet.