# Colouring amino acid substitutions by scoring matrix

I have a list of site-specific amino acid substitutions from a mutagenesis experiment which I am visualising by colouring the structure of the protein in PyMOL. I would like a way to colour the substitutions by some approximation of the biochemical (dis)similarity between them.

So far, I have tried to create a colour gradient of scores from an alignment scoring matrix (e.g. BLOSUM62, therefore a 16-shade white to red gradient; a substitution of A to R for example gets a -1). However, I am fairly certain this is not the best interpretation of of the log-odds scores in the BLOSUM matrix, and I wondered if anyone had any thoughts on how I might better accomplish this?

Note that other colour schemes such as hydrophobicity, Tailor etc. aren't suitable for my purpose.

• If you wanted to plot the diversity of an MSA Cosurf (consurf.tau.ac.il) would the tool to go for (consurfDB is the precalculated version). For a set of point mutations and using PyMOL only (ouch!), I would suggest checking this PyMOL snippet, which gives a rough ∆∆G, which could be set as a b factor (alter resi xx, b=1.23) as played with that way. Oct 11 at 13:12
• *) I mention ∆∆G because that would be a metric of change in stability. Whereas a substitution matrix is a metric of how uncommon such a change is from a selection point of view. If your experiment is a Lenski style experiment (long term evolution experiment), that makes sense, but not if it's directed evolution following in vitro mutagenesis (mutazyme, magnanese, wobbly analogues etc) even if with in vivo selection (e.g. phage display, biosensor etc etc.) as the mutational spectrum is a product of the in vitro step not biological selection of a population. Oct 11 at 13:19
• Hi @MatteoFerla many thanks for your thoughts. I think the Gibbs energy approach is maybe not quite what I am looking for because we know that both the start and the end point fold and are active - we are particularly interested in IDing important substitutions for changes in activity between the two isoforms (this being a somewhat unguided experiment, based on ancestral sequence reconstruction) Oct 11 at 13:27
• Ah. That makes sense if it's an ASR —plus you probably have some rad epistatic mutations, which the ∆∆G of each in isolation would have been dismal. A substitution matrix is poor metric for dissimilarity because it would just light up the surface while in reality the slow evolving core mutations are the interesting ones. Consurf would reflect this, but is not specific for your changes. Is surface/core the reason why hydrophobicity, volume etc etc. are not suitable? Oct 11 at 14:06
• Indeed. So ideally we would like some way to winnow this (obviously other than exhaustive combinatorial mutagenesis, which we have in the pipeline). Oct 11 at 17:24