There are significant challenges faced in the field of drug discovery, the accurate prediction of bioactive molecules’ interaction with a respective biological target is one of them. Latent Tuberculosis Infection (LTBI) affects approximately 1.7 billion people. Individuals with LTBI have a 5–10% lifetime risk of developing active TB, and current treatments for LTBI show 60–90% efficacy. With drug resistance increasing, the WHO recommends systematic testing and treatment of LTBI in high-risk populations. Dihydrofolate reductase (DHFR) is a key enzyme in Mycobacterium tuberculosis (Mtb) that plays an essential role in DNA replication and repair. Inhibiting DHFR disrupts the folate pathway, making it a promising target for anti-tubercular drug discovery. The proposed research focuses on identifying active inhibitors of the DHFR enzyme using Quantitative Structure–Activity Relationship (QSAR) modeling. The approach includes generating molecular descriptors, developing a QSAR model, and applying various machine learning models to predict the bioactivity of molecules against DHFR using pIC50 values. Unlike previous studies that typically use Random Forest Regressor, this research compares 42 regression models through the Lazy Predict module, selecting the optimal model based on R2 and RMSE scores. Feature selection was performed using the variance threshold method with a threshold of 0.16 to enhance computational speed and reduce overfitting. Gradient Boosting Regressor, after hyperparameter tuning, yielded an R2 score of 0.40 and an RMSE of 1.24. This systematic approach improves drug discovery efficiency by accurately predicting potential DHFR inhibitors, offering insights into combating drug-resistant LTBI.