Skip to main content

Table 3 Prediction Accuracy of naive Knn methods and RB-KNN methods. A combination of data sources is represented by a binary string, in which each bit specifies whether a data source is included, in the order of expression correlation, chromosomal distance, block indicator, and paralog indicator. glm refers the RBKNN method based on logistic regression, and loess refers to the RBKNN method based on local regression, loess is not applicable to "0001" (paralog indicator) and "0010" (block indicator) since the only data source is binary.

From: A Regression-based K nearest neighbor algorithm for gene function prediction from heterogeneous data

Combination

naive knn

glm

loess

0001

0.41

0.57

NA

0010

0.41

0.54

NA

0100

0.47

0.47

0.56

1000

0.40

0.44

0.52

1010

0.52

0.50

0.55

1011

0.61

0.68

0.68

1111

0.52

0.68

0.68