Description:
Regression model to predict inhibitors of glycogen phosphorylase b (GPB). The model was built with the Multiple Linear Regression technique by using a total of 8 QuBiLS-MAS descriptors.

Training and testing datasets:
A total of 44 training compounds and 22 testing compounds were extracted from the Sutherland, et al., 10.1021/jm0497141

Internal performance:
For a 10-fold cross-validation repeated 100 times: Squared R = 0.8052, MAE = 0.3872, and RMSE = 0.4765.

External performance:
Squared R = 0.7700, MAE = 0.4611, and RMSE = 0.601.

Regression equation:
pKi =

      0.3509 * AC[5]_MX_Q_AB_nCi_2_SS10_H_P_LGP[5-6]_dc4_MAS +
      4.1478 * GV[3]_N3_B_AB_nCi_2_MP13_H_n_T_LGP[1-2]_e-p_MAS +
     -0.2272 * GV[6]_S_B_AB_nCi_2_SS14_H_C_LGP[2-3]_c-dc4_MAS +
    944.4828 * GV[2]_AM_Q_AB_nCi_2_MP13_C_LGP[2-4]_est_MAS +
     -0.1127 * ES_K_B_AB_nCi_2_SS15_n_C_LGP[4-5]_psa-hx_MAS +
     -8.6556 * GV[4]_N3_Q_AB_nCi_2_MP14_n_C_LGP[3-6]_li_MAS +
      0.3822 * AC[2]_S_Q_AB_nCi_2_SS2_H_A_LGP[2-7]_ec_MAS +
      0.0245 * TS[4]_SD_Q_AB_nCi_2_SS15_H_D_LGP[1-4]_alk_MAS +
     -0.3498