Ligand Prediction for Orphan Targets Using Support Vector Machines and Various Target-Ligand Kernels Is Dominated by Nearest Neighbor Effects

Abstract

Support vector machine (SVM) calculations combining protein and small molecule information have been applied to identify ligands for simulated orphan targets (i.e., targets for which no ligands were available). The combination of protein and ligand information was facilitated through the design of target-ligand kernel functions that account for pairwise ligand and target similarity. The design and biological information content of such kernel functions was expected to play a major role for target-directed ligand prediction. Therefore, a variety of target-ligand kernels were implemented to capture different types of target information including sequence, secondary structure, tertiary structure, biophysical properties, ontologies, or structural taxonomy. These kernels were tested in ligand predictions for simulated orphan targets in two target protein systems characterized by the presence of different intertarget relationships. Surprisingly, although there were targetand set-specific differences in prediction rates for alternative target-ligand kernels, the performance of these kernels was overall similar and also similar to SVM linear combinations. Test calculations designed to better understand possible reasons for these observations revealed that ligand information provided by nearest neighbors of orphan targets significantly influenced SVM performance, much more so than the inclusion of protein information. As long as ligands of closely related neighbors of orphan targets were available for SVM learning, orphan target ligands could be well predicted, regardless of the type and sophistication of the kernel function that was used. These findings suggest simplified strategies for SVM-based ligand prediction for orphan targets.

CompoundProtein Interaction Prediction Within Chemogenomics: Theoretical Concepts, Practical Usage, and Future Directions

“…[42] In a followup study, Wassermann, Geppert, and Bajorath examined the effect of ligands predicted for orphan targets in which closely related proteins of the target orphan had a set of ligands. [43] In such cases, they concluded that the ligands of the related proteins were more influential than particular optimizations of interaction prediction schemes. The result Special Issue Chemogenomics Table 1.…”

Machine learning and similarity-based virtual screening techniques

“…the design of combined ligand-target kernels. A variety of kernel functions accounting for target sequence, structure or ontology information at different levels of sophistication have been combined with ligand similarity kernels to examine whether SVM LBVS performance might be further increased [17]. However, this was often not the case, revealing that compound similarity and nearest neighbor effects mostly dominated the recall rates of SVM calculations [17].…”

Structure‐Based Approaches to Target Fishing and Ligand Profiling

“…A recent work form Bajorath and colleagues [99] suggests that simplified strategies for designing target-ligand kernels should be used since varying the complexity of the target kernel does not influence much the identification of ligands for virtually deorphanized targets. However, no true 3-D cavity descriptor has yet been reported as target kernel.…”