Predicting Proteome-Early Drug Induced Cardiac Toxicity Relationships (Pro-EDICToRs) with Node Overlapping Parameters (NOPs) of a new class of Blood Mass-Spectra graphs

Blood Serum Proteome-Mass Spectra (SP-MS) may allow detecting Proteome-Early Drug Induced Cardiac Toxicity Relationships (called here Pro-EDICToRs). However, due to the thousands of proteins in the SP identifying general Pro-EDICToRs patterns instead of a single protein marker may represents a more realistic alternative. In this sense, first we introduced a novel Cartesian 2D spectrum graph for SP-MS. Next, we introduced the graph node-overlapping parameters (nop_k) to numerically characterize SP-MS using them as inputs to seek a Quantitative Proteome-Toxicity Relationship (QPTR) classifier for Pro-EDICToRs with accuracy higher than 80%. Principal Component Analysis (PCA) on the nop_k values present in the QPTR model explains with one factor (F₁) the 82.7% of variance. Next, these nop_k values were used to construct by the first time a Pro-EDICToRs Complex Network having nodes (samples) linked by edges (similarity between two samples). We compared the topology of two sub-networks (cardiac toxicity and control samples); finding extreme relative differences for the re-linking (P) and Zagreb (M2) indices (9.5 and 54.2 % respectively) out of 11 parameters. We also compared subnetworks with well known ideal random networks including Barabasi-Albert, Kleinberg Small World, Erdos-Renyi, and Epsstein Power Law models. Finally, we proposed Partial Order (PO) schemes of the 115 samples based on LDA-probabilities, F₁-scores and/or network node degrees. PCA-CN and LDA-PCA based POs with Tanimoto’s coefficients equal or higher than 0.75 are promising for the study of Pro-EDICToRs. These results shows that simple QPTRs models based on MS graph numerical parameters are an interesting tool for proteome research.