1 reply [Last post]
jeff101's picture
User offline. Last seen 10 hours 19 min ago. Offline
Joined: 04/20/2012
Groups: Go Science

When you mutate amino acids in a protein, there are many things to consider:

(1) Odds are not all mutations are equally likely.
I wouldn't expect there to be a 1 in 20 chance to mutate to ala
and a 1 in 20 chance to mutate to trp.

(2) Different types of proteins have different distributions of amino acids in them.
Consider eukaryotic vs prokaryotic, membrane vs cytosolic proteins, etc.

(3) Some amino acids have more different DNA/RNA codons than others.
For example, https://en.wikipedia.org/wiki/DNA_codon_table lists
1 codon for trp and 6 codons for leu. From this, you'd expect
mutations to leu to be 6x more likely than mutations to trp.

You could also consider that changing one base in a codon can
change which amino acid it codes for. One could envision a chart
that connects a starting amino acid and all its possible codons to
other amino acids with codons that differ by only one base. In this
chart, some amino acids will connect to each other more often than

It might even be that certain bases are more likely to change than others.
For example, A->C might happen more often than G->C or A->T.
Perhaps even the base's position within the codon affects mutation rates.

(4) Some amino acids are more likely to be on the protein surface
while others are more likely to be in the protein core. If you sort the
amino acids by their present locations in the protein (surface vs core),
you can pick which amino acid mutations are more likely to work.

(5) You can classify amino acids by size, polarity, and charge.
See https://en.wikipedia.org/wiki/Amino_acid#/media/File:Amino_Acids.svg for example.
Mutating to fairly similar amino acids should make minor changes to the protein structure.

For example, mutating phe -> tyr, lys -> arg, his -> trp, thr -> ser -> cys,
asp -> asn -> gln -> glu -> asp, and gly -> ala -> val -> iso -> leu
all look like fairly gradual structural transitions.

Also, replacing positively-charged ones (arg, his, lys) with each other,
negatively-charged ones (asp, glu) with each other, polar uncharged ones
(ser, thr, asn, gln) with each other, or hydrophobic ones
(ala, val, iso, leu, met, phe, tyr, trp) with each other all seem like
reasonable amino acid mutations.

Do you know of any Foldit Mutate Recipes that account for these things?

jeff101's picture
User offline. Last seen 10 hours 19 min ago. Offline
Joined: 04/20/2012
Groups: Go Science
Different secondary structures contain different amino acids:

(6) Some amino acids are more likely to be found in helices,
some are more likely in sheets, and others are more likely in loops.
The Chou-Fasman method uses this idea to predict secondary structures
but it could be used to pick mutations as well. For example,
if you are mutating a helix, try changing to helix-forming residues
like ala, glu, leu, or met, and avoid helix-ending ones like pro or gly.
If you are mutating a sheet, try changing to sheet-forming residues.
If you are mutating a loop, try changing to residues like pro and gly,
which are often found in turns.


Developed by: UW Center for Game Science, UW Institute for Protein Design, Northeastern University, Vanderbilt University Meiler Lab, UC Davis
Supported by: DARPA, NSF, NIH, HHMI, Amazon, Microsoft, Adobe, Boehringer Ingelheim, RosettaCommons