Somatic mutation detection using ensemble of flexible neural tree model

The advances on next-generation sequencing technology (NGS) have enabled researchers to detect somatic mutations. Much effect has been devoted to improve accuracy of discovering somatic mutations from tumour/normal NGS data. In this study, flexible neural tree model (FNT) is proposed to detect somatic mutations in tumour-normal paired sequencing data. To improve the classification accuracy further, a new classification ensemble approach based on Radial Basis Function (RBF) neural networks as nonlinear combination function is proposed. The proposed method is applied to real biological dataset from exome capture data and the whole genome shotgun data. Results show that the obtained FNT model has a fewer number of variables with reduced number of input features and with significant improvement in the detection accuracy using the proposed ensemble learning method. Our method also selects 10 import features for somatic mutation detection, which could be used to analyze NGS mutations further.