Toward the prediction of FBPase inhibitory activity using chemoinformatic methods

Area of Science:

• Chemoinformatics
• Computational Chemistry
• Drug Discovery

Background:

• Fructose 1,6-bisphosphatase (FBPase) is a key enzyme in gluconeogenesis, making its inhibitors potential therapeutic agents for type 2 diabetes mellitus.
• Predictive chemoinformatic models are crucial for identifying novel FBPase inhibitors efficiently.

Purpose of the Study:

• To develop and validate a robust computational model for predicting FBPase inhibitory activity.
• To identify potential oxazole and thiazole analogs as FBPase inhibitors for type 2 diabetes treatment.

Main Methods:

• A genetic algorithm-random forest (GA-RF) coupled method was employed using Mold(2) molecular descriptors.
• The GA-RF model was trained on 126 oxazole and thiazole analogs and validated on an independent set of 64 compounds.
• Model performance was assessed using correlation coefficients (r(2) ncv, r(2) cv, r(2) pred) and validated against established criteria (r(2) o, r(2) m).

Main Results:

• The GA-RF model demonstrated high predictive accuracy with r(2) ncv = 0.96, r(2) cv = 0.67, and r(2) pred = 0.90.
• The model successfully met validation criteria with r(2) o = 0.90 and r(2) m = 0.83.
• The GA-RF model outperformed a pure random forest model in terms of prediction capacity.

Conclusions:

• The developed GA-RF model possesses significant internal and external prediction capabilities for FBPase inhibitors.
• This computational approach can accelerate the drug discovery process by prioritizing potential oxazole and thiazole FBPase inhibitors for synthesis and testing.