Quantitative Structure-Activity Relationship (QSAR) is a computational approach that correlates chemical structure with biological activity using molecular descriptors and statistical models. For B. Pharm students, QSAR is essential for rational drug design, lead optimization, and ADME/Tox prediction. Key topics include descriptor types (topological, electronic, hydrophobic, geometrical), model-building methods (MLR, PLS, SVM, random forest), validation strategies (cross-validation, external validation, Y‑randomization), and applicability domain. Mastery of descriptor calculation, feature selection, alignment in 3D-QSAR, and interpretability helps predict activity and reduce experimental costs. Now let’s test your knowledge with 30 MCQs on this topic.
Q1. What does QSAR primarily relate to biological activity?
- Molecular descriptors derived from chemical structure
- Clinical trial outcomes
- Manufacturer batch numbers
- Physician prescribing habits
Correct Answer: Molecular descriptors derived from chemical structure
Q2. Which descriptor type captures hydrophobic character important for membrane permeability?
- Topological descriptors
- Electronic descriptors
- Hydrophobic (logP) descriptors
- Geometrical descriptors
Correct Answer: Hydrophobic (logP) descriptors
Q3. What is the purpose of splitting data into training and test sets in QSAR?
- To increase descriptor dimensionality
- To evaluate model generalizability on unseen data
- To remove outliers from the dataset
- To normalize descriptor scales
Correct Answer: To evaluate model generalizability on unseen data
Q4. Which statistical method is commonly used for simple, interpretable QSAR models?
- Partial least squares (PLS)
- Multiple linear regression (MLR)
- Convolutional neural networks (CNN)
- K-means clustering
Correct Answer: Multiple linear regression (MLR)
Q5. What does a high q2 (cross-validated R2) value indicate?
- Poor internal predictivity
- Good internal predictive ability of the model
- Overfitting to the training data
- Large applicability domain
Correct Answer: Good internal predictive ability of the model
Q6. Which technique is specific to 3D-QSAR and visualizes steric and electrostatic fields?
- HQSAR
- CoMFA (Comparative Molecular Field Analysis)
- MLR
- Topological analysis
Correct Answer: CoMFA (Comparative Molecular Field Analysis)
Q7. What is Y‑randomization used to test in QSAR modeling?
- Descriptor calculation accuracy
- Whether the model performance is due to chance correlations
- Computational speed of the algorithm
- Applicability domain size
Correct Answer: Whether the model performance is due to chance correlations
Q8. Which descriptor category includes molecular weight and atom counts?
- Electronic descriptors
- Geometrical descriptors
- Constitutional descriptors
- Topological descriptors
Correct Answer: Constitutional descriptors
Q9. What problem arises when descriptors are highly collinear?
- Improved model interpretability
- Unstable coefficient estimates and multicollinearity issues
- Increased external predictivity
- Reduced descriptor count automatically
Correct Answer: Unstable coefficient estimates and multicollinearity issues
Q10. Which feature selection method searches descriptor space using evolution-inspired operators?
- Stepwise regression
- Genetic algorithm (GA)
- Principal component analysis (PCA)
- Hierarchical clustering
Correct Answer: Genetic algorithm (GA)
Q11. In QSAR, the applicability domain defines:
- The software license terms
- The chemical space where model predictions are reliable
- The type of machine used for computation
- The number of descriptors used
Correct Answer: The chemical space where model predictions are reliable
Q12. Which validation assesses predictive power on completely unseen external molecules?
- Internal cross-validation (leave-one-out)
- External validation using a test set
- Descriptor scaling
- Descriptor pruning
Correct Answer: External validation using a test set
Q13. What is the main advantage of PLS over MLR in QSAR?
- PLS cannot handle collinearity
- PLS reduces dimensionality and handles multicollinearity well
- PLS always produces simpler models than MLR
- PLS does not require descriptor calculation
Correct Answer: PLS reduces dimensionality and handles multicollinearity well
Q14. Which of the following is a 2D-QSAR method that uses fragments as descriptors?
- CoMSIA
- HQSAR (Hologram QSAR)
- CoMFA
- 3D field mapping
Correct Answer: HQSAR (Hologram QSAR)
Q15. Which metric measures average magnitude of prediction errors (lower is better)?
- r2
- q2
- RMSE (Root Mean Square Error)
- Descriptor variance
Correct Answer: RMSE (Root Mean Square Error)
Q16. Which preprocessing step helps make descriptors comparable by scale?
- Y‑randomization
- Descriptor normalization or standardization
- External validation
- Grid spacing selection
Correct Answer: Descriptor normalization or standardization
Q17. In 3D-QSAR, why is molecular alignment important?
- Alignment defines comparative positions of molecules for field calculations
- Alignment increases computation time without benefit
- Alignment eliminates need for descriptors
- Alignment is only used for 2D-QSAR
Correct Answer: Alignment defines comparative positions of molecules for field calculations
Q18. Which machine-learning method is non-linear and useful for complex QSAR patterns?
- Multiple linear regression (MLR)
- Support vector machine (SVM)
- Stepwise regression
- Simple averaging
Correct Answer: Support vector machine (SVM)
Q19. What does an r2 value close to 1 indicate for a QSAR model on training data?
- Perfect external predictivity always
- Good fit to the training data
- Model has no descriptors
- Applicability domain is infinite
Correct Answer: Good fit to the training data
Q20. Which descriptor type captures electronic distribution like partial charges?
- Hydrophobic descriptors
- Electronic descriptors
- Constitutional descriptors
- Topological indices
Correct Answer: Electronic descriptors
Q21. What is a common sign of model overfitting?
- High training r2 but low external predictive performance
- Low training r2 and high test performance
- Balanced training and test performance
- Small number of descriptors
Correct Answer: High training r2 but low external predictive performance
Q22. CoMSIA differs from CoMFA by:
- Using hologram fragments
- Comparing only 2D descriptors
- Using Gaussian-type functions for similarity fields including hydrophobic and H-bond descriptors
- Being identical in all procedures
Correct Answer: Using Gaussian-type functions for similarity fields including hydrophobic and H-bond descriptors
Q23. Which approach helps interpret which molecular features increase activity?
- Random descriptor removal
- Contour maps from 3D-QSAR and coefficient interpretation in MLR/PLS
- Using larger training sets only
- Ignoring applicability domain
Correct Answer: Contour maps from 3D-QSAR and coefficient interpretation in MLR/PLS
Q24. Which performance measure evaluates how much variance is explained by the model?
- RMSE
- r2 (coefficient of determination)
- Descriptor count
- Grid spacing
Correct Answer: r2 (coefficient of determination)
Q25. What is the role of an applicability domain (AD) check before using a QSAR prediction?
- To assess whether the compound lies in model’s reliable chemical space
- To compute r2 automatically
- To remove descriptors that are invalid
- To convert 3D structures to 2D
Correct Answer: To assess whether the compound lies in model’s reliable chemical space
Q26. Which descriptor is a topological index representing molecular branching?
- LogP
- Wiener index
- Partial charge
- Polar surface area (PSA)
Correct Answer: Wiener index
Q27. Why is external validation preferred over only internal cross-validation?
- External validation is faster
- External validation better assesses true predictive ability on independent data
- Internal cross-validation always overestimates error
- External validation eliminates the need for descriptors
Correct Answer: External validation better assesses true predictive ability on independent data
Q28. Which QSAR workflow step directly follows descriptor calculation?
- Model deployment to production
- Feature selection and preprocessing
- Clinical trials
- Grid spacing optimization
Correct Answer: Feature selection and preprocessing
Q29. What is one benefit of consensus modeling in QSAR?
- It always produces the simplest model
- Combining predictions from multiple models can improve robustness and reduce error
- It eliminates need for validation
- It reduces descriptor diversity
Correct Answer: Combining predictions from multiple models can improve robustness and reduce error
Q30. Which QSAR practice enhances mechanistic interpretability of models?
- Using many correlated descriptors without reporting coefficients
- Prioritizing interpretable descriptors, visualizing contour maps, and reporting coefficients
- Never validating the model externally
- Avoiding reporting of applicability domain
Correct Answer: Prioritizing interpretable descriptors, visualizing contour maps, and reporting coefficients
