Pharmacoinformatics
Pharmacoinformatics in NIPER JEE covers cheminformatics, molecular modeling, docking, QSAR, bioinformatics databases, and ADMET prediction.
Core rules and concepts
- Molecular descriptors include MW, logP, TPSA, rotatable bonds, HBD/HBA, and fingerprints. Lipinski and Veber rules filter oral candidates.
- Similarity: Tanimoto coefficient = common bits/(A + B − common bits). Values near 1 mean high similarity; activity cliffs can break similarity–activity assumptions.
- Docking predicts poses and scores; scoring functions approximate binding affinity and are not free-energy measurements. Redocking a crystallographic ligand validates the protocol.
- QSAR needs training/test splitting, internal cross-validation, external validation, and an applicability domain. Overfitting gives good training R² but poor prediction.
- Databases: PubChem (compounds), ChEMBL (bioactivity), PDB (structures), DrugBank (drugs), UniProt (proteins). Homology modeling needs a template with sufficient sequence identity.
Worked example
Fingerprint A has 12 bits, B has 9 bits, and 6 bits are common. Tanimoto = 6/(12 + 9 − 6) = 6/15 = 0.40. This moderate similarity suggests related chemistry but does not prove shared activity; a bioassay or QSAR model is still required.
Traps that cost marks
- Reporting a docking score as a measured ΔG.
- Assuming high Tanimoto similarity guarantees the same target activity.
- Using R² alone to validate QSAR; external prediction is essential.
- Building a homology model on a low-identity template and treating it as experimental.