The effects of shared information on semantic calculations in the gene ontology

Abstract

The structured vocabulary that describes gene function, the gene ontology (GO), serves as a powerful tool in biological research. One application of GO in computational biology calculates semantic similarity between two concepts to make inferences about the functional similarity of genes. A class of term similarity algorithms explicitly calculates the shared information (SI) between concepts then substitutes this calculation into traditional term similarity measures such as Resnik, Lin, and Jiang-Conrath. Alternative SI approaches, when combined with ontology choice and term similarity type, lead to many gene-to-gene similarity measures. No thorough investigation has been made into the behavior, complexity, and performance of semantic methods derived from distinct SI approaches. We apply bootstrapping to compare the generalized performance of 57 gene-to-gene semantic measures across six benchmarks. Considering the number of measures, we additionally evaluate whether these methods can be leveraged through ensemble machine learning to improve prediction performance. Results showed that the choice of ontology type most strongly influenced performance across all evaluations. Combining measures into an ensemble classifier reduces cross-validation error beyond any individual measure for protein interaction prediction. This improvement resulted from information gained through the combination of ontology types as ensemble methods within each GO type offered no improvement. These results demonstrate that multiple SI measures can be leveraged for machine learning tasks such as automated gene function prediction by incorporating methods from across the ontologies. To facilitate future research in this area, we developed the GO Graph Tool Kit (GGTK), an open source C++ library with Python interface (github.com/paulbible/ggtk).

An Integrated Platform Supporting Semantic Similarity Score Calculation and Reproducibility

“…Retrieving, reproducing and reusing SS scores for any ontology in any application is still challenging 39,44,54 . This mainly due to the lack of a tool that exhaustively implements existing SS models and related assumptions to produce consistent scores on demand and in real-time for use in related applications and for testing hypotheses.…”

“…Symbols of different measures used in some existing tools and PySML are shown in Supplementary File 1 (see Table 1) and we refer readers to Supplementary File 2 (see Appendix 2) where complete descriptions and algebraic forms of all these measures are provided. Best performance often results from trading between accuracy and computational speed, PySML implements all these measures, except those which are known to be computationally unattractive with high disproportion between computational complexity and performance improvement, e.g., measures built on graph-based similarity measure (GRASM) 12,54 . However, PySML offers a platform that may be used to easily develop, test and assess any measures, so that users who are interested in these measures, for example, can process them within the platform.…”

“…The high scores produced by SML is mainly caused by the contribution of the root of the ontology in SS computation as proteins sharing only the ontology root have a score greater than 0. This suggests that SML overestimates SS scores by considering the root as an informative ontology concept, which ultimately biases these scores 38 , thus negatively impacting the performance of these SS models 38,54 .…”

An improved approach to infer protein-protein interaction based on a hierarchical vector space model

“…The effects of the shared information for the semantic similarity calculation were discussed in [ 41 ]. The shared information of a term pair is the common inheritance relations extracted from the structure of the GO graph.…”

“…Experiments of three different methods calculating the term similarity, each with five shared information methods, were done on three ontologies across six benchmarks. Among the choice of shared information, term similarity algorithm, and ontology type, the choice of ontology type most strongly influenced the performance, and shared information type had the least influence [ 41 ].…”