Estimation of protein-compound interaction and rational design of compound library based on chemical genomic information

A non-transitory computer-readable medium containing programming instructions that cause a computer processor to perform a method for estimation of protein-compound interaction of chemical substances, wherein the chemical substances include:

• a first chemical substance, wherein the first chemical substance is a small molecule and comprises a first characteristic amount expressed as a first vector having a compound descriptor calculated from the structure and properties of the first chemical substance, wherein the structure and properties of the first chemical substance are expressed as a plurality of compound descriptors including constitutional descriptors, topological descriptors, walk and path counts, connectivity indices, information indices, 2D autocorrelations, edge adjacency indices, burden eigenvalue descriptors, topological charge indices, eigenvalue-based indices, functional group counts, atom-centered fragments, and molecular properties;
• and
• a second chemical substance, wherein the second chemical substance is a protein and comprises a second characteristic amount expressed as a second vector having a protein descriptor calculated from only an amino acid sequence of the second chemical substance;
• the method being characterized by comprising:
• (A) constructing a learning model using a Support Vector Machine algorithm to obtain a hyperplane for performing pattern recognitions of one or more type of information of the first chemical substance and of the second chemical substance, both of which are to be estimated for protein-compound interaction, wherein the first characteristic amount of the first chemical substance having the compound descriptor expressed as the first vector and the second characteristic amount of the second chemical substance having the protein descriptor expressed as the second vector are map- transformed to maximize correlation between chemical space coordinate and biological space coordinate,
• (B) mapping a characteristic vector constructed by combining the first characteristic amount of the first chemical substance to be estimated for interaction and the second characteristic amount of the second chemical substance to be estimated for interaction, into the learning model constructed in step (A) for pattern recognitions, and
• (C) determining whether or not the first chemical substance to be estimated for interaction interacts with the second chemical substance to be estimated for interaction based on a mapped-position of the characteristic vector in step (B) relative to the hyperplane obtained in step (A).