Background: Heat Shock Protein 90(HSP90) inhibitors are involved in multiple anticancer
pathways, which indicate many important novel molecular targets for cancer therapy. However,
the characteristics of poor water solubility, liver toxicity and finite bioavailability of the present
inhibitors limit clinical application. Hence, it is crucial to evaluate the characteristics of compounds
and develop new drugs with hypotoxicity and high-bioactivity.
Methods: Quantitative Structure-Activity Relationship (QSAR) has been an effective method for
screening novel structures and predicting various properties of the synthesized compounds. Heuristic
Method (HM) and Gene Expression Programming (GEP) algorithm were used to establish linear and
nonlinear models severally.
Results: The results showed that HM has good correlation coefficients of R2 and lower S2 as 0.79
and 0.29 for the training set and GEP has better values of 0.89 and 0.05, respectively.
Conclusion: Both models have the capability of prediction but the nonlinear model developed by
GEP has a more excellent predictive ability and indicates further optimization of the HSP90 inhibitors.