A data processing method for an estimation of compound-protein interaction using both chemical substance information, such as a chemical property of the compound, and biological information, such as sequence information of genes to rationally and efficiently screen compounds. First space representing space coordinates of a first chemical substance group and second space representing space coordinates of a second chemical substance group are defined, and the first chemical substance group is characterized by a first characteristic amount and the second chemical substance group is characterized by a second characteristic amount, and map transformation of the coordinates of the first space and the coordinates of the second space results in the solution so as to increase the correlation between the first space and the second space using a multivariable analysis technique or a machine learning method.