G-Protein-Coupled Receptors (GPCRs) are cell membrane proteins of relevance to biology and pharmacology. Their supervised classification in subtypes is hampered by label noise, which stems from a combination of expert knowledge limitations and lack of clear correspondence between labels and different representations of the protein primary sequences. In this brief study, we describe a systematic approach to the analysis of GPCR misclassifications using Support Vector Machines and use it to assist the discovery of database labeling quality problems and investigate the extent to which GPCR sequence physicochemical transformations reflect GPCR subtype labeling. The proposed approach could enable a filtering approach to the label noise problem.

Peer Reviewed