Unsupervised machine learning, QSAR modelling and web tool development for streamlining the lead identification...

Area of Science:

• Computational Chemistry
• Medicinal Chemistry
• Drug Discovery

Background:

• Traditional lead compound identification is costly and slow.
• In silico methods offer a promising alternative for drug discovery.
• Antimalarial drug development requires efficient lead identification strategies.

Purpose of the Study:

• To develop predictive 2D-QSAR models for identifying potential antimalarial flavonoids.
• To predict the inhibitory concentration 50 (IC50) against Plasmodium falciparum.
• To create a sustainable in silico approach for antimalarial drug development.

Main Methods:

• Machine learning algorithms (PCA, K-means clustering) and Pearson correlation were used.
• Nine molecular descriptors were selected for quantitative structure-activity relationship (QSAR) model building.
• Multiple linear regression (MLR) was applied 100 times to select and validate the best QSAR models.

Main Results:

• Three robust and predictive 2D-QSAR models were developed and validated.
• The models successfully predicted IC50 values against Plasmodium falciparum.
• The developed models adhere to the OECD principles for QSAR validation.

Conclusions:

• A reliable and sustainable in silico method for predicting antimalarial activity of flavonoids was established.
• The study significantly reduces the time and cost associated with identifying antimalarial lead compounds.
• A web tool, JazQSAR, was developed for accessible use of the QSAR models.