search

ilia-kats / NRPSDesigner

design non-ribosomal peptide synthases from scratch
3 stars 3 forks source link

Epimerisation domains #51

Open nignatiadis opened 10 years ago

nignatiadis commented 10 years ago

NRP: L-2,4-diaminobutyric acid + L-alanine Output: A-T-C_L-A-T-E-TE with the second adenylation domain coming from Tolypoclatium inflatum, simA with specificity for D-alanine.

Two issues here: 1) Are we in general able to incorporate Epimerisation domains just before TE domains (any biological evidence)? In all other cases we get the appropriate specificity after an E domain by action of C domains..

2) We also have A-domains in the database with speficity for L-alanine (B.subtilis FenE7A), thus there's no reason to do it via D-alanine+Epimerisation. To avoid this, I suggest to give a high penalty to "T-E" type combinations, or even do this manually in the designer code?..

ilia-kats commented 10 years ago

Apparently penicillin synthetases have an eprimerization domain preceding a TE domain, see http://www.sciencedirect.com/science/article/pii/S0300908411002926 . So I guess the question is whether we want to penalize evolutionary distance or domains involved more, and if penalizing domains involved, up to which evolutionary distance.

Hetitus commented 10 years ago

I don't know if I got it correctly but I would say that if there is know other domain available with the right specificity then just include the penalized domain. Maybe we can also display the penalty of a designed contract visually somewhere in the gui - then the user would know about the quality of the construct...

nignatiadis commented 10 years ago

@Hetitus ^ problem was that there exists domain with right specificity (A domain for L-alanine) but dijkstra returns 2 domains (A domain for D-alanine + Epimerisation) instead of correct one.

@ilia-kats Ok the paper seems interesting (will read later) and we should probably open a new issue for thioesterase domains to discuss this further.. In regards to A/E, I think I don't like this possibility at all (by going A+E instead of only A you add 1 kb to the construct without any reason..), so maybe this should already be done before dijkstra, when the graph is constructed? As in:

Search for A domains. If they exist, add them to the graph. Only if they do not exist search for A domains for the enantiomer and add to graph.

What do you think?

ilia-kats commented 10 years ago

Question is whether a construct containing only A, but with a high evolutionary distance to the rest of the construct, is more likely to work than a construct containig A+E where the whole construct is assembled from closely related sources.