Abstract
e16523 Background: IARC classified arsenic (As) as “carcinogenic to humans”, but despite the health consequences, there is no molecular signature available yet to predict when exposure may lead to the disease development. First aim of this study is to investigate the genetic changes due to the exposure to carcinogenic compounds such as As. Secondly, how accurately we can predict the disease association when exposed to toxic compounds. Methods: The entire analysis was performed in-silico fashion and data was collected from the public resources such as NCBI database. Two Asian population datasets exposed As were used to find significantly differently express genes. In addition, four cancer cell lines with exposure of As compounds were used to identify the association with cancer. The human bladder cancer biopsy datasets were used to develop a risk predictive model. As per the requirements, numerous machine learning (ML) approaches such as random forest, hierarchical clustering, were used to find the classification and association between the samples and outcome. Statistical approaches such as T-Test, and ANOVA, applied to find the differentially expressed genes associated with different conditions and logistic regression models applied to develop risk prediction models. Results: We identified a set of 1183 genes which were common between both the populations and were significantly changed in humans exposed to As. A subset of 157 genes associated with As exposure and involved in cancer progression was selected for risk prediction model development. A set of four genes (NKIRAS2, AKTIP, HLA-DQA1 and TBC1D7) shows the highest prediction ability of primary bladder tumor with AUC 0.96 (95% CI: 0.82- 0.99) and reproducibility of 0.75 (0.34-0.93) when applied on different dataset. Conclusions: This study identified a list of genes and bladder cancer predictive models that would be very helpful to forecast the outcomes of As exposed in humans.
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have