Two-stage ML Classifier for Identifying Host Protein Targets of the Dengue Protease.

Overview

Flaviviruses such as dengue encode a protease that is essential for viral replication. The protease functions by cleaving well-conserved positions in the viral polyprotein. In addition to the viral polyprotein, the dengue protease cleaves at least one host protein involved in immune response. This raises the question, what other host proteins are targeted and cleaved? Here we present a new computational method for identifying putative host protein targets of the dengue virus protease. Our method relies on biochemical and secondary structure features at the known cleavage sites in the viral polyprotein in a two-stage classification process to identify putative cleavage targets. The accuracy of our predictions scaled inversely with evolutionary distance when we applied it to the known cleavage sites of several other flaviviruses-a good indication of the validity of our predictions. Ultimately, our classifier identified 257 human protein sites possessing both a similar target motif and accessible local structure. These proteins are promising candidates for further investigation. As the number of viral sequences expands, our method could be adopted to predict host targets of other flaviviruses.