Predicting potential residues associated with lung cancer using deep neural network

Abstract

Lung cancer is a prominent type of cancer, which leads to high mortality rate worldwide. The major lung cancers lung adenocarcinoma (LUAD) and lung squamous carcinoma (LUSC) occur mainly due to somatic driver mutations in proteins and screening of such mutations is often cost and time intensive. Hence, in the present study, we systematically analyzed the preferred residues, residues pairs and motifs of 4172 disease prone sites in 195 proteins and compared with 4137 neutral sites. We observed that the motifs LG, QF and TST are preferred in disease prone sites whereas GK, KA and ISL are predominant in neutral sites. In addition, Gly, Asp, Glu, Gln and Trp are preferred in disease prone sites whereas, Ile, Val, Lys, Asn and Phe are preferred in neutral sites. Further, utilizing deep neural networks, we have developed a method for predicting disease prone sites with amino acid sequence based features such as physicochemical properties, conservation scores, secondary structure and di and tri-peptide motifs. The model is able to predict the disease prone sites at an accuracy of 81 % with sensitivity, specificity and AUC of 82 %, 78 % and 0.91, respectively, on 10-fold cross-validation. When the model was tested with a set of 417 disease-causing and 413 neutral sites, we obtained an accuracy and AUC of 80 % and 0.89, respectively. We suggest that our method can serve as an effective method to identify the disease causing and neutral sites in lung cancer. We have developed a web server CanProSite for identifying the disease prone sites and it is freely available at-https://web.iitm.ac.in/bioinfo2/CanProSite/.