QSPR model for predicting bioavailability of orally-taken anti-cancer drugs / 6
6
Melchizedec G. Garcia.
-
-
- 82 pp. 28 cm.
-
-
-
-
- .
- .
- 0 .
- .
- 0 .
Thesis: (Bachelor of Science in Chemistry) - Pamantasan ng Lungsod ng Maynila, 2022.
5
ABSTRACT: The bioavailability data of fifty-three orally-taken anti-cancer compounds were subjected to quantitative structure-property relationship analysis to create a predictive model. Three-hundred sixty-five physicochemical, 224 topological, and 189 ligfilter descriptors were screened and analysed using multiple linear regression to identify descriptors with the highest contribution to drug bioavailability. A total of 600 MLR analyses with varying numbers of y-variable were done throughout the study and only a single model was able to produce satisfactory results. Descriptors included in this model are aasC_Key, sOH_Sum, sOH_Avg, Sum of topological distances between F..I, Topplogical diameter, ALOGP7, Maximal electro topological positive variation, Second Mohar, Valence connectivity index chi-3, and Num aliphatic rings. To test the model's stability, robustness, and predictive power, it was subjected to internal and external validation. In both validation processes, the model performed well based on the statistical criteria set in the literature. (Internal validation: RMSE=0.11, R=0.83, R2=0.69, R2adj=0.60. R2-R2-R2adj=0.09, and Q2=69; External Validation: Q2=0.73, R2=0.88, (R2=0.73, R2=0.88, (R2-Ro2)/R2=0.08, R2-R02pred=0.99, k=1.04, K1=0.96).