I wish to investigate how a virus could be attenuated through targeted mutagenesis by a priori prediction of attenuating amino acid mutations.

In particular, I am seeking bioinformatic solutions where the protein sequence of an exisiting live vaccine strais has been compared against it wild-type counterparts to identify key attenuating mutations and the key attenuating mutation(s) can be deduced without wet-lab site directed mutagenesis.

I presume surface antigens, such as the spike protein in COVID-10 would be the proteins of interest, but I'm open to suggestions.

I am a computer scientist looking to investigate the mutational parameter space of viral protein sequence data. A key approach I am considering is to use a mutational correlation/covariance matrix and then remove convolution artefacts. I hope this will identify compensatory mutations that will maintain a local interaction for example within a fold, providing a starting point for identifying attenuating mutations.

This may have application for example to COVID-19 and the goal is to produce an efficacious vaccine.

  • $\begingroup$ COVID-19 is a complicated genome and the starting point is to get to grips with its fundamental biology. I strongly disagree with the statement bioinformatics has no role to play - quite the contrary you will not get anywhere without it right now. $\endgroup$
    – M__
    Mar 7 '20 at 15:33

There isn't a vaccine for any coronavirus, and your question is generally about targeted attentuation, which is a complex area.

The basic building blocks for any vaccine development is virological understanding of the proteins involved in pathogenesis. I will focus on covid-19 as an example here.

The majority of bioinformatics work is based around the spike protein as is wetlab vaccine development. This gene is immediately after the 20 kb of ORF1ab, including the RNA dependent RNA polymerase, its around 3.5 kb but it might be a bit bigger. The spike protein is a major B-cell (antibody) antigenic determinant and the cellular receptor sites are located within it. Your bioinformatics could ultimately focus on targeting these specific sites, in particular using antigenic (antibody) prediction algorithms.

First you need some basic biology of spike and hydrophobicity prediction is a good starting point such as TMHMM here and consider the SARS spike protein structure for example here. Once you've a handle on those structures antigenicity scanning would be a next phase analysis and possibly homology modelling. This is because COVID-19 and SARS are closely related, but there are key differences. You finally consider the full range of amino acid mutations in current epidemic via NCBI.

At this point you would target a specific vaccine design strategy and have a clearer understanding on the approach you wish to take. Good luck!


Not the answer you're looking for? Browse other questions tagged or ask your own question.