Integrated Molecular Modelling and Machine Learning Strategies to Investigate Potential Drug Targets Against Poliovirus

Abstract

Poliovirus is a highly pathogenic virus causing the crippling disease of Poliomyelitis. Poliovirus mostly infects infants with a weak immune system and is still infecting many in developing countries like Pakistan and Afghanistan. Poliovirus virus drug designing efforts might help in the eradication of poliovirus. Poliovirus is a non-enveloped +ssRNA virus that can cause paralytic polio by the chromatolysis of the motor neurons residing in the spinal cord or brain stem. The non-structure proteins of poliovirus are involved in the proteolysis of the single polypeptide into functional proteins. These are the cysteine proteases i.e., 2A and 3C proteases. A previous proof of concept study identifies poliovirus protease 3C and 2A also play a crucial role in the apoptosis of the motor neuron. The 2A and 3C protease initiate apoptosis by caspase-independent and dependent pathways respectively. The 2A protease also cleaves the nuclear pore proteins Nup62, Nup98, and EIF4G1. This blocks the transport of host mRNA required for the viability of cells and results in the nuclear localization of protease 3C. Additionally, 3C protease degrades the DNA and cleaves the poly (ADP-ribose) involve in DNA repair. The 3C protease also cleaves cytoskeletal protein MAP4 and translocates cytochrome c from mitochondria. These morphological changes by the 3C protease induce apoptosis by activation of the caspase pathway. Moreover, protease 2A and 3C cleaves the protein Eukaryotic translation initiation factor 4 G (eIF4G) and Poly(A)-binding protein (PAB or PABP) which terminate the translation of the host cells. This involvement of 2A and 3c proteases makes them a significant target for a drug against poliovirus. This study aims to explore 2A and 3C protease as a potential drug target against poliovirus. For this purpose, approaches like homology modeling and MD simulation have been used to get a stable molecular structure of proteases. The docking experiments have been used to probe the best binding confirmation of the ligands with the proteases. The MD simulation of some ligand complexes was performed to evaluate the ligand-protein interaction profiles. One of the challenges faced while targeting the viral proteases is the conserved nature of the viral proteases with human proteases. This highly similarity of viral proteases with humans can lead to the offtarget toxicity which can be controlled by increasing the specificity of drug towards the viral proteases. In order to identify the unique classification features of viral x proteases we have used machine learning technique of decision tree. Additionally, ensemble methods like the random forest, bagging, and boasting have been used to remove the bias and variance in the data. These strategies identified that hydrogen bonding is the most crucial interaction required for the inhibition of 2A and 3C protease. Moreover, the residues Cys147 and Gln146 have displayed stable interaction in more than one complex of 3C and 2A protease respectively. The machine learning techniques highlights sequence features like the length of the sequence, frequency of proline, and alanine as the most significant feature for the classification of viral protease from human protease. This project explores the poliovirus conserved proteases 2A and 3C as therapeutic targets which might help in antiviral drug development.