Query on NonSilent and Silent mutations..

[shyam_la]

What is the ratio of non-silent to silent mutations resulting from single nucleotide substitutions that can be expected, if every base is equally mutable and there is no selection pressure?

Hope the question is clearly framed.

I am asking so that I can comment on whether my dataset deviates from this hypothetical ratio and hence an idea of how much "Selection" really occurred.

[BAMseek]

I wrote some code that basically goes through all 64 possible nucleotide triplets, looks at every possible 1-base substitution (9 possibilities each), and checks if that substitution changes the amino acid.

The code for this is available at https://gist.github.com/4162316

I get 576 total substitutions (64*9). 138 are silent and 438 are nonsilent. This would be a ratio of silent:nonsilent = 138:438, which is approx. 1:3.17

Does that make sense? This was done pre-coffee, so let me know if you find some issues.

Justin

[HESmith]

Shyam La,

You'll also need to factor in the frequencies of the different types of mutations (e.g., transitions vs transversions) produced by your mutagen, as most exhibit significant biases.

[shyam_la]

Justin,

That is exactly what I wanted to know; mutate each codon and get the result. Thank you!
But I'm guessing 1:3.17 is not a biologically meaningful ratio because typically cancers (which is what I am dealing with) have something like 1:2 and they also are subjected to some amount of selection pressure which means we would expect a ratio bigger than 1:3.17 (more non silent than silent).

HESmith's idea that mutagen biases have to be included is a valid argument, I guess..

I wrote some code that basically goes through all 64 possible nucleotide triplets, looks at every possible 1-base substitution (9 possibilities each), and checks if that substitution changes the amino acid.

The code for this is available at https://gist.github.com/4162316

I get 576 total substitutions (64*9). 138 are silent and 438 are nonsilent. This would be a ratio of silent:nonsilent = 138:438, which is approx. 1:3.17

Does that make sense? This was done pre-coffee, so let me know if you find some issues.

Justin

[BAMseek]

Yeah, that model might be too simplistic, but I guess it gives a baseline considering all things equal.

From what I've seen in the literature, the transition to transversion ratio is around 2-2.1 for the entire genome, and around 2.8-3.0 within the exome.

A more realistic approach might be to take a bunch of normal human exomes and calculate the silent to non silent mutations. Or, instead of trying to find silent/non-silent, maybe calculate the transitions to transversions in your dataset and see if that agrees with the numbers found in literature.

Justin

[mlodato]

Perhaps this is useless since the thread is so old, but I have been thinking about this as well and came across this paper:

PMID:16783027

which seems to model SNS as well as regional biases to get expected numbers of missense vs nonsense mutations.