Is there a tool to predict if a missense variant causes a loss of function on the protein?

Question

[giusdalt95]

Good morning, Does anybody know wether there is a tool to predict the effect (loss or gain of function) of a missense mutation on a protein? I have already checked that my mutations are pathogenic according to PolyPhen-2, SIFT, MCAP13, CADD and MutationTaster; furthermore, I checked the effect of the aminoacidic substitutions on my proteins with tools like SNAP2 and DUET. Nonetheless, I looked up the literature and found that apparently these tools do not seem to be able to discriminate between a loss or a gain of function mutation.

Thank you in advance, Giuseppe

[karl.stamm]

Interested if you find one. I don't think it's well defined. We can predict if the protein will be significantly different, but whether that's good or bad depends on the whole animal and environment.

Answer 1

[Jorge Amigo]

https://pcingola.github.io/SnpEff/se_inputoutput/#loss-of-function-lof-and-nonsense-mediated-decay-nmd-predictions

[Collin]

I don't think the link is relevant for missense mutations as specified by the poster. Missense mutations can be loss-of-function without truncating the protein.

Answer 2

[Collin]

As some one who does a lot of variant interpretation in cancer genomes, a priori it is more likely the variant is loss-of-function than gain-of-function. This is for two reasons: 1) pathogenicity predictors are typically better at predicting pathogenic loss-of-function mutations (you indicated that they were predicted by various methods); 2) typically it is easier to lose protein function than to gain/improve one. That being said, gain-of-function pathogenic mutations are quite important in disease. Usually one of the best indicators that a mutation is gain-of-function is seeing multiple independent pathogenic mutations in a very localized area in the protein. In cancer genomics these are called mutational "hotspots". Pathogenic germline mutations are also highly localized for autosomal dominant (AD) diseases (https://pubmed.ncbi.nlm.nih.gov/26246501/ ), likely because AD inherited diseases are more likely to be caused by gain-of-function pathogenic mutations. However, "dominant negative" mutations can be an exception to the rule-of-thumb that multiple independent pathogenic mutations that are clustered together are gain-of-function.

As a side note, I have previously looked at how well we can distinguish between pathogenic mutations in oncogenes (gain-of-function) versus tumor suppressor genes (loss-of-function) (https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4930736/ ). The distinction was reasonable but not perfect. Mutations in oncogenes (gain-of-function) tended to occur over smaller regions of the protein, be more solvent exposed, less likely to be at hydrophobic residues, and tended to be at residues that are more evolutionarily conserved compared to mutations in tumor suppressor genes.

Answer 3

[Mensur Dlakic]

Nonetheless, I looked up the literature and found that apparently these tools do not seem to be able to discriminate between a loss or a gain of function mutation.

The answer to this question depends on how much we already know about the mutated residue. For a random residue in a random protein, it is very difficult to make a solid prediction. On the other hand, if we know that a residue is catalytic, or part of the binding interface with other biological molecules, it is easier to make a meaningful prediction.

Most prediction tools rationalize how the mutation will behave with regard to solvent accessibility (exposed vs. buried), amino-acid properties (large vs. small, charged vs. neutral, polar vs. non-polar, etc) or how it fits the protein backbone in terms of energy (assuming that we know protein's 3D structure). The thing with proteins is that not all of their regions are equally susceptible to mutations, and they generally don't fall apart regardless of mutations. Not to be morbid, but consider a person losing a finger. It makes a big difference whether they lost a finger fully or partially, which finger and on what hand. Similarly, there are only so many cases where mutations change protein functions in ways that are either black or white - in most cases there are shades of gray.