Aritra Mahapatra and Jayanta Mukherjee *
Background: Advancement in the sequencing technology yields a huge number of genomes of a multitude of organisms in our planet. One of the fundamental tasks for processing and analyzing these sequences is to organize them in the existing taxonomic orders.
Method: Recently we proposed a novel approach, GenFooT, of taxonomy classification using the concept of genomic footprint (GFP). The technique is further refined and enhanced in this work leading to improved accuracies in the task of taxonomic classification on various benchmark datasets. GenFooT maps a genome sequence in a 2D coordinate space and extracts features from that representation. It uses two hyper-parameters, namely block size and number of fragments of genomic sequence while computing the feature. In this work, we propose an analysis for choosing values of those parameters adaptively from the sequences. The enhanced version of GenFooT is named GenFooT2.
Results and Conclusion: We have experimented GenFooT2 on ten different biological datasets of genomic sequences of various organisms belonging to different taxonomy ranks. Our experimental results indicate more than 3% improved classification performance of the proposed features with Logistic regression classifier than the GenFooT. We also performed the statistical test to compare the performance of GenFooT2 with the state-of-the-art methods including our previous method GenFooT. The experimental results as well as the statistical test exhibit that the performance of the proposed GenFooT2 is significantly better.
Taxonomy classification, Mitochondrial genome, Genomic footprint, Alignment-free, Shannon entropy, Supervised classification, Biological dataset
Department of Computer Science and Engineering, Indian Institute of Technology Kharagpur, Department of Computer Science and Engineering, Indian Institute of Technology Kharagpur