Computational Approach for Homology Discovery of Keratin Digestion Genes in Zebrafish

Abstract

Abstract Computational approaches for gene prediction have drawn a significant importance considering the pace at which raw sequences of biological data is getting available in past few decades where biological experiments for drawing the meaningful insights from these raw data have failed to meet this pace. This research study focuses on gene prediction towards the functionality of keratin digestion in scale eating. For this gene prediction, genomic data of zebrafish is used against the known keratin digestion data of keratin-feeding clothes moths and keratin beetles. Since fishes and insects are highly different organisms, it created the requirement to build a comprehensive pipeline for the gene prediction. Hence we first clustered the Expressed Sequence Tags (ESTs) and then they were passed through a motif discovery process. As the next step, those motifs were matched against the genome of zebrafish by performing a homology search. Exhibiting promising results, we could achieve a match hit with an E - value of 0.058 that starts at the location of 14411 bp in the genome of zebrafish. To further evaluate the obtained match, a requirement to develop a model that can claim whether a given sequence is a gene or not was raised. As such, in the next phase of the pipeline, a Markov model for CpG island prediction was designed and developed and that model successfully shows an accuracy of 93.5%. Finally, we passed the starting region of the obtained match to this model and most importantly, the model predicted it as a CpG island. This suggests that the obtained match exhibits the properties of a gene which can be considered as the ultimate highest goal that can be achieved in a computational gene prediction research. Keywords: Gene Prediction, Homology Discovery, Keratin, Lepidophagy, CpG Island Prediction, Markov Models