# Gene Expression, Docking and Machine Learning in Malaria Drug Discovery: A Systematic Review

## TL;DR

This review explores how gene expression, molecular docking, and machine learning are being used to discover new malaria drugs, focusing on herbal compounds and computational methods.

## Contribution

The paper systematically reviews the integration of molecular and computational techniques for antimalarial drug discovery from 2014 to 2024.

## Key findings

- Molecular docking was the most used technique, with compounds like isorhamnetin showing strong binding to Plasmodium proteins.
- ML models predicted bioactivity and resistance patterns, identifying flavonoids and terpenoids as promising drug candidates.
- RNA-seq analysis revealed key genes modulated by herbal treatments, including those involved in apoptosis and inflammation.

## Abstract

Malaria remains a significant and worldwide health threat with increasing resistance to current treatments, stimulating the demand for innovative approaches in pursuing drug discovery. This systematic review integrates the progress made from 2014 through 2024 regarding molecular methods like gene expression profiling, molecular docking and machine learning to understand the biology of Plasmodium and identify new drug targets and compounds, focusing on herbal remedies and computational methods.

Several studies were found using a PRISMA‐guided search of PubMed, Scopus and Web of Science (64 studies found). The data extracted were gene expression outcomes, docking affinities, ML models and experimental validations (in vitro/in vivo).

Molecular docking emerged as the dominant technique (32.37%), followed by in vitro antiplasmodial assays (14.39%), ADMET profiling (10.79%) and gene expression studies (3.60%). RNA‐seq analysis revealed key host and parasite genes modulated by herbal treatments, including those involved in apoptosis and inflammation. Notably, compounds like isorhamnetin and myricetin 3‐O‐glucoside showed exceptionally high binding affinities to Plasmepsin II and Plasmodium falciparum lactate dehydrogenase (PfLDH) (ΔG < −13 kcal/mol). ML models like random forest and support vector machine (SVM) exhibited high predictive results (AUC value up to 0.87) for bioactivity and resistance patterns that showed flavonoids (quercetin) and terpenoids (eugenol) as good candidates. Pathways that are often attacked are haemoglobin degradation, glycolysis, pyrimidine metabolism and protein synthesis.

Multiomics, docking and ML integration improve the target identification and prioritise the compounds. This review illustrates the great potential of molecular techniques for the development of drugs against antimalarial helicases that are not resistant to drug therapy. However, in vivo data holes and methodology inconsistency limit clinical translation. Future work should include standardisation of protocols and studies of synergistic combinations of phytochemicals.