Closed marco-brandizi closed 1 year ago
I remember we tested this variant after publishing the paper to reduce the amplitude of the score while maintaining the order of the gene rank. I don't think it has an impact on the actual gene ranking (order of genes). The main factor that impacts the ranking is the number of evidences. Leave this as is for now until we find time to fully revise and enhance the scoring.
While working on the new API, I've noticed this, here:
double evidenceWeight = ( igf + luceneScore + invGraphDistance ) / 3;
These are the three terms described in the papers, but shouldn't they be multiplied? Aren't them too apples and oranges to make an average?