Quantitative Structure-Activity Relationship (QSAR) modelling of the activity of anti-colorectal cancer agents featuring quantum chemical predictors and interaction terms

•Logistic regression was used to obtain coefficients for the QSAR model.•Total electronic energy and most positive atomic charge as predictors.•Interaction terms between predictors are significant.•Aromatic carbons, carbonyl, ArNHR and oxetane fragments constitute active compounds. A Quantitative Structure-Activity Relationship (QSAR) study was performed to 1) predict the activity of new compounds as anti-colorectal cancer agents and 2) select chemical predictors that accurately model relationships between molecular structures and bioactivity against colon cancer cell lines. Logistic regression was used to obtain coefficients for the QSAR classification model. Quantum chemical predictors were found to be significant predictors, such as the total electronic energy (ET), charge of the most positive atom (Qmax) and electrophilicity (ω), thus validating the need to consider them besides those routinely used in anti-colorectal cancer agents QSAR. The model selected simple and significant predictors that can be obtained directly using information from gaseous-state Gaussian optimisation at HF/3-21G in a vacuum, with no external calculation software, providing an inexpensive yet robust QSAR model. In addition, two interaction terms were proven to be significant at 95 % confidence level, reiterating the importance of interaction between predictors that is not commonly seen in QSAR models.