studying the genetic loci related to the bovine leukemia virus using random forest method and genomic data

Bovine leukemia virus (blv) is a causative agent of bovine leukosis, which, due to its long incubation period, can spread widely within a herd before clinical symptoms appear, causing significant economic losses. this study used a supervised machine learning method called random forest to identify genomic regions associated with blv. the non-parametric nature of this method allows for the creation of predictive models without the need for initial statistical assumptions; whereas the standard genome-wide association studies (gwas) methods are usually based on single-variable hypothesis tests and cannot account for correlations resulting from connectivity imbalance or the combination of multiple markers. in this study, the genotyping data of 145 holstein cows (77 blv-positive, 68 healthy) after quality control by using the plink (v 1.02), which resulted in 23,910 single nucleotide polymorphisms (snps) were analyzed. random forest analyses on the mentioned data included three hyperparameters: mtry (0.5(p/3), (p/3), 2(p/3)), ntree (2000, 3000, 4000), and nodesize (5, 10, 15), where p is equal to the total number of snps (23,910). to find the best snps, the mean decrease accuracy (mda) index (> 1.89) was used which resulted in the selection of 50 snps. genomic enrichment analyses showed that genes associated with the top 50 snps are predominantly involved in positive regulation, intracellular signaling, apoptosis and cell death, signal transduction, metabolic processes, and cell differentiation and development. in total, 82 genes were identified, including hub genes such as myc, rabif, irs1, trappc9, mapk8, htt, snx9, bclaf1, xrn1, and lsm6.