Exomiser
Exomiser copied to clipboard
exomiser
missense tolerance ratio (MTR)
missense tolerance ratio (MTR) uses GNOMAD data to assess whether certain regions of proteins are susceptible to disease-causing missense variants.
- http://mtr-viewer.mdhs.unimelb.edu.au/
- https://genome.cshlp.org/content/27/10/1715
Assess whether this score will improve the performance of Exomiser with missense variants, maybe by the following plan
- take the current Exomiser missense score of max(MutationTaster, PolyPhen, SIFT)
- multiply the score by the MTR
- Rescale to keep final scores in [0,1]
- Then, we can calculate the precision/recall or ROC for all of the missense mutations in ClinVar or GeL cases. Note that these scores are not used to calculate the MTR.
- If we see a substantial boost in the AUROC or area under the PR curve, then we could add this as an option or default for Exomiser.
From Slave Petrovski (now in Cambridge, UK) ftp://mtr-viewer.mdhs.unimelb.edu.au for full download
Within that FTP are two exome-wide datasets depending on your needs: MTR flat file: MTR score for each possible missense variant (ie. by genomic co-ordinates (hg19) and ref/alt), per transcript MTR table: MTR score for each amino acid position, per transcript
The above two are 'transcript-aware', if you prefer a single one then there's also the canonical-only version as a separate file that contains the set of canonical ENST's (v75) you can filter by. These were manually selected as being the canonical-defined tx in gnomAD, and if none assigned, the longest available transcript for the gene symbol.
We also have b38 version of MTR if you need; just let me know.. Excited to see how it performs and if it adds additional predictive utility to your existing scores.. We have found it to be quite striking in our hands, fingers-crossed you have similar experience.. Also, happy to collaborate if opportunities arise.
I am wondering if the easiest thing would be to calculate the MTR ourselves using the Exomiser database, which would avoid various conversion issues such as having the MTR at the protein but not the nucleotide level?
