Cross-validation and cross-study validation of chronic lymphocytic leukaemia with exome sequences and machine learning

The era of genomics brings the potential of better DNA-based risk prediction and treatment. We explore this problem for chronic lymphocytic leukaemia that is one of the largest whole exome data set available from the NIH dbGaP database. We perform a standard next-generation sequence procedure to obtain Single-Nucleotide Polymorphism (SNP) variants and obtain a peak mean accuracy of 82% in our cross-validation study. We also cross-validate an Affymetrix 6.0 genome-wide association study of the same samples where we find a peak accuracy of 57%. We then perform a cross-study validation with exome samples from other studies in the NIH dbGaP database serving as the external data set. There we obtain an accuracy of 70% with top Pearson ranked SNPs obtained from the original exome data set. Our study shows that even with a small sample size we can obtain moderate to high accuracy with exome sequences, which is encouraging for future work.