A deep learning model-based personalizedmedicine approach to preventing drug resistance in ALKpositive non-small cell lung cancer (NSCLC) patients

Background: The most common type of lung cancer, nonsmallcell lung cancer (NSCLC), can be caused by mutations inthe ALK gene, which result in resistance to ALK inhibitors andrecurrence of the disease. ALK gene mutations can be identifiedand predicted using deep learning. By combining artificial intelligencemodeling with dynamic drug inhibition mechanisms,personalized medicine can take a new step forward.Materials and Methods: A mutation has been identified in differentpositions of the ALK gene according to the COSMICdatabase. Combining these mutations with deep learning enabledus to identify possible subsequent mutations at differentlevels. Two mechanisms are considered in the model of inhibitionof drug action following mutation: a change in release energycaused by the EML۴-ALK fusion, as well as the geometricposition of the TKI at the time of activation. Logistic regressionswere used to determine the correlation between drug andgenetic models.Results: There was an increase in release energy in ۲۰۱ patientswith mutations in the ALK gene following mutations inthe L۱۹۹۶M position. There is a significant correlation betweenthe different mutation sites and the maximum geometric matchof TKIs and the activation of c_Met ۴۸ in resistant patients.Discussion: The increase in energy released is due to the lossof competitive effect of the ATP binding pocket following thedecrease in the activity of ALK inhibitor drugs. Based on themodeling, it was found that the geometric pairing of the tyrosinekinase with its inhibitor was more accurate. Due to personalizedmedicine, it is possible to use next-generation drugs thatinhibit ALK according to an individual's gene mutations. It hasbeen observed that mutations in multiple positions have led tochanges in both pathways of drug action. Therefore, it seemsthat multi-gene modeling can be used to predict the specificdrug that is suitable for each patient.