| Title: Determining confidence of predicted interactions between HIV-1 and human proteins using conformal method |
| Abstract: Identifying protein-protein interactions (PPI's) is critical for understanding virtually all cellular molecular mechanisms. Previously, predicting PPI.s was treated as a binary classification task and has commonly been solved in a supervised setting which requires a positive labeled set of known PPI's and a negative labeled set of non-interacting protein pairs. In those methods, the learner provides the likelihood of the predicted interaction, but without a confidence level associated with each prediction. Here, we apply a conformal prediction framework to obtain confidence levels of predictions without relying on a negative dataset. The conformal predictor uses a function measuring relative "strangeness" of pairs to check whether an unknown prediction added to the sequence of already known PPI's would conform to the "exchangeability" assumption which reasonably assumes that the interactions are normally being discovered in random order. The method suffers less from artifacts related to the over-representation of some proteins in the dataset used for training. The guarantee of finding a particular number of known interactions at a given confidence level helps the biologist in the interpretation of the results, and better assists the choices of pairs for experimental validation. We apply the proposed conformal framework to improve the identification of interacting pairs between HIV-1 and human proteins. |
Paper: PDF |
Supplement Materials: Zipped Files |
PSB 2012: Talk Slide |