Open RachelKarchin opened 5 months ago
From Rachel - Here is a paper for people to read before the session: https://pubmed.ncbi.nlm.nih.gov/38956207/
A preliminary survey of interest (where every participant was allowed to vote twice) resulted in 10 votes for this topic.
LInk to Rachel's presentation: https://www.dropbox.com/scl/fi/psu8t1inz1iqq7t3l73xz/CGC-Unconference-Transparency.pptx?rlkey=trd11jhxaxu5bn6ubg1g0wo6h&dl=0
Link to "Are Next-Generation Pathogenicity Predictors Applicable to Cancer?" from Anna Pachenko's lab (Journal Molecular Biology 2024) https://www.dropbox.com/scl/fi/2mgnqn99xj5of0p70o57z/Are-Next-Generation-Pathogenicity-Predictors-Applicable-to-Cancer-Panchenko.pdf?rlkey=wq3jyocb1gdmvp08itof3hxzf&dl=0
Link to CHASMplus paper (Cell Systems 2019) https://www.dropbox.com/scl/fi/kjzvdnt5v186ffhpzqyc1/Tokheim-Cell-Systems-2019.pdf?rlkey=7njekqanas5z4brr2jy0sqven&dl=0
Submitter Name
Kyle Moad/Rachel Karchin
Submitter Affiliation
Johns Hopkins
Submitter Github Handle
kmoad/RachelKarchin
Additional Submitter Details
We are the PI and an engineer from the team behind OpenCRAVAT, a meta-annotation software framework, for variant interpretation. OpenCRAVAT provides predictions from over 30 variant effect prediction tools. We want to increase the utility and transparency of these predictions.
Project Details
We will discuss issues surrounding the growing integration of computational tools for variant effect prediction, into the diagnostic process. These include developing approaches to allow users to interpret and reason about predictions, to make sense of diverse predictions from multiple predictors, and to map prediction scores to the ACMG/AMP/VICC recommendations for classification of germline and somatic variants. Specific topics include: evidence double counting; the necessity for transparency in the features and logic underpinning predictions; understanding training data to prevent circular reasoning, and how to interpret results from increasingly popular 'black box' AI models. Aimed at clinicians, diagnostic personnel, and anyone interested in the future of genomic medicine, this workshop promises to provide valuable insights into improving the reliability of variant pathogenicity classifications and fostering the clinical application of predictive methods, ultimately advancing patient care in the realm of personalized medicine.
Required Knowledge
Familiarity with genetics/genomics and an interest in variant effect prediction.