SQUIRLS

[damiansm]

• Peter's team is implementing an algorithm to detect exonic splicing enhancer mutations. I think we can try to do that, cryptic splicing, and MaxEnt analysis of the entire canonical splice site (not just +1,+2,-1,-2). This should immediately improve the diagnostic yield by 10-20% if the software halfway works.
• Also assess whether scoring the first and last position in each exon highly improves things. Do biocuration of positive literature cases for training. Should be highly likely to affect splicing. Note these are a subset of the variants labelled SO SPLICE_REGION_VARIANT by Jannovar: the 3 exonic base at end of exon and 5 intronic bases 3' of the splice donor OR the 5 intronic bases 5' of the splice acceptor and 3 exonic bases at start of exon
• Assess performance on GEL cases
• See #133 as well
• Daniel and Peter have been working on this and the ESE approaches
• We discussed with Jules adding code to Exomiser to retrieve xxbp around a potential splice variant of interest that will the input into Daniel's code and return a variant score from 0 to 1. Peter can collaborate to confirm any predictions via miniGene constructs. Can test on GEL cases. https://github.com/TheJacksonLaboratory/ESElator/blob/development/eselator-bass/src/main/java/org/monarchinitiative/bass/BassAnnotator.java (development branch) This is the basic function that will coordinate the analysis. Daniel has implemented all of the relevant parts of the pipeline incluyding the Rogan score, MaxEnt etc, and we have also implemented a score for the exonic splice enhancer mutations. BassAnnotator(BassSource bassSource, JannovarData jannovarData) It will not need Jannovar in the Genomiser version; BASS source will have things like the exon susceptibility score, knowledge of whether an exon is the first or last etc, and could be implemented by hooking up to the H2 database We will implement this function to return the most pathogenic score from these cases depending on where the mutation is in the exon/intron. Daniel has also implemented a framework that will let us test this prior to integration in the Genomiser.

[damiansm]

Now published as SQUIRLS: https://pubmed.ncbi.nlm.nih.gov/34289339/

Presumably we can now close https://github.com/exomiser/Exomiser/issues/133 and just focus on how to incorporate SQUIRLS into Exomiser analysis. I have successfully generated a tabix score files for all SNVs that can be run as a local source.