BLKnn: A K-nearest neighbors method for predicting bioluminescent proteins

Bioluminescence is a chemical process in which light is produced and emitted by a living organism. Recent biotechnological applications of bioluminescence include using of bioluminescent proteins in gene expression analysis, bioluminescent imaging, study of protein-protein interaction and disease progression, drug discovery, toxicity determination, etc. Therefore, it is of great medical and commercial significances to identify bioluminescent proteins accurately and efficiently. In this study, we present BLKnn, a K-nearest neighbors method that can predict bioluminescent proteins. This method is based on the bit-score weighted Euclidean distance, which is calculated from compositions of selected amino acids and pseudo-amino acids. On a balanced training dataset, BLKnn achieved 74.9% sensitivity, 95.5% specificity, 85.2% accuracy, and 0.919 AUC (area under the ROC curve) by 10-fold cross-validation. When tested on a much bigger independent test dataset, the method also achieved a consistent performance of 88.0% overall accuracy and 0.989 AUC. Comparisons showed that BLKnn outperformed previously published methods. The method is available at https://edisk.fandm.edu/jing.hu/blknn/blknn.html.