We have worked during the last years on assessing the functional impact of non-synonymous variants (nsSNVs). As a result, we have published two new approaches Condel and transFIC. In this post I would like to clarify the differences between one and the other, and give our recommendations on when each of them should be used.
The development of Condel was motivated by the observation that different tools for the prediction of variants involved in disease (eg. SIFT, PolyPhen2, Mutation Assessor, etc) often gave different results, although all of them had in general quite good performance in differentiating disease mutations versus polymorphisms.
With this observation in mind we developed a method to combine the output of various tools with the idea that this combined score could improve the performance of each method alone in the prediction of disease variants. We tested several approaches and we found that the weighted average of the scores of 5 methods outperformed significantly each of the methods alone (see ROC curve). The weights are derived from the benchmark of the tools in a dataset of disease versus polymorphisms.
Note that Condel was benchmarked with two datasets of disease versus polymorphisms, and the weights assigned to each tool are derived from similar datasets. Thus we recommend the use of Condel to assess the functional impact (deleteriousness or disease implication) of germline variants. You can obtain Condel scores using Condel web server.
Condel has been used to assess the functional impact of cancer somatic mutations, and we showed that cancer recurrent mutations have a high Condel score and that the Condel score of TP53 mutations correlates with TP53 protein activity. These are indications that Condel is useful to assess the impact of cancer somatic mutations, although it was not developed with the aim of ranking cancer somatic mutations.
TransFIC (transformed Functional Impact for Cancer)
The development of transFIC was motivated by the reasoning that the functional impact of mutations in cancer cells would depend not only on how the mutation affects the function of the protein, but also on how critical is the protein for cell functioning. To quantify protein criticality for the cell systematically, we derived a measure of the tolerance of proteins to functional impact mutations in the human population (baseline tolerance). Next, we used this measure to transform the scores of well known methods to assess the functional impact of mutations (eg. SIFT, PolyPhen2 and Mutation Assessor) and we called this transformation transFIC. We tested the performance of the transFIC of these scores versus the original scores using nine proxy datasets of cancer driver and passenger mutations (read more about the evaluation approach). We found that the transFIC improves significantly the capacity of those methods to differentiate likely drivers from likely passengers (read more here). In particular, the transFIC of Mutation Assessor (MA) is the one that performs better, thus we recommend the use of the transFIC of MA to prioritize somatic mutations for their implication in cancer. You can compute this score using transFIC webserver.
In summary, our recommendations are:
- If you want to prioritize germline variants for their functional impact (deleteriousness or disease implication), use Condel.
- If you want to prioritize cancer somatic variants for their possible implication in tumor development, use transFIC.
Gonzalez-Perez A and Lopez-Bigas N. Improving the assessment of the outcome of non-synonymous SNVs with a Consensus deleteriousness score (Condel). Am J Hum Genet 2011, doi:10.1016/j.ajhg.2011.03.004.
Gonzalez-Perez A, Deu-Pons J and Lopez-Bigas. Improving the prediction of the functional impact of cancer mutations by baseline tolerance transformation. Genome Medicine 2012. 4:89 doi:10.1186/gm390s