Figure 1.

Overview of evolutionary chemical binding similarity (ECBS) method. (A) The simplified chemical-target-evolutionary relationship is used as an example for a schematic description of different ECBS models. The ‘C’ in the first layer means a chemical compound, the ‘T’ means a target, and the ‘F’ represents arbitrary evolutionary information (e.g. family). The C-T connection is defined by the direct binding between C and T, and T-F is defined by the evolutionary class information F of the target T. The chemical pairs are classified based on the evolutionary relationship of binding targets. For example, C1–C2, C3–C4 and C5–C6 are evolutionarily related chemical pairs (ERCPs) by the common binding targets, and C1–C3, C1–C4, C2–C3 and C2–C4 are also ERCPs defined by F1. The other unrelated chemical pairs are considered as negative data. (B) The conceptual classification scheme for different ECBS models is depicted using the example in (A). In the QSAR-like model, active molecules are defined for the target T1 and the classification model (dotted line) clusters them together in chemical feature space. In the Target-ECBS model, the ERCPs defined by the multiple targets (T1, T2 and T3) will be clustered by the model, but the ERCPs by T1 and T2 are likely to have closer distances in chemical feature space because of the similar molecular features for binding to the evolutionarily conserved binding pocket defined by F1. On the other hand, the Family-ECBS model starts to consider the ERCPs defined by evolutionary information of the targets (chemical pairs in the green background). These additional ERCPs will make an enhancement effect to locate the F1-related chemical compounds in close proximity by presenting evolutionarily conserved features more evidently. The TS-Family-ECBS model is a target-specific Family-ECBS model that only considers the ERCPs defined from the targets evolutionarily related to the predefined target T1 (square box). Therefore, the model construction procedure is identical to the Family-ECBS model except that C5–C6 is excluded because the target T3 has no evolutionary relationship to T1 or T2. Compared to the target-specific model, the Target-ECBS and Family-ECBS are categorized as a unified model, because it considers multiple and heterogeneous chemical-target binding information altogether.