Genome Music Smg: How Can A Single Mutation Be Statistically Significant?

Question

[Christian]

I am going through my list of significantly mutated genes as determined by 'genome music smg' and am puzzled by the fact that many of the genes in this list carry a single mutation in a single patient and yet have assigned a highly significant p-value. How is this explained? Isn't it a neccessity that a gene is mutated multiple times (either within one patient or between patients) such that a p-value can be meaningfully computed?

Here is one example of such a gene, taken from the output of the smg analysis:

#Gene    Indels    SNVs    Tot Muts    Covd Bps    Muts pMbp    P-value FCPT    P-value LRT    P-value CT    FDR FCPT    FDR LRT    FDR CT
LMOD1    0    1    1    55    18181.82    0.000661479    4.91E-07    4.86E-11    0.487266065    0.000280645    1.24E-07

Answer 1

[Chris Miller]

MuSiC's SMG test takes both the background mutation rate and the gene size into consideration as it asks the question: "is this gene mutated more frequently than we'd expect by chance?" If the BMR is very low and the gene in question is very small, then even a single mutation can occasionally be significant. That said, you should treat this just like any other statistical test and remember that statistically significant does not necessarily imply importance in cancer (though it may hint at it). It's only one piece of information that should be taken into consideration.

[Cyriac Kandoth]

This is correct. Christian, in your example, notice how LMOD1 only has 55 covered bps across all your samples. You might want to loosen the minimum tumor/normal read depths, in the calc-covg step before running calc-bmr and subsequently the SMG test.