Lecture 7: QSAR and Drug Development Flashcards
QSAR (Quantitative Structure-Activity Relationship)
-math models showing correlation between drug structure and actiivties
-derived by statistical regression
-information on receptor not necessary
Molecular Descriptors
-numerical parameters describing chem properties that can be directly determined from structure
-used as variables in QSAR models
Molecular descriptors examples
-drug as whole: LogP and D, molecular weight, pKa
-fragments: pi and sig values, size
Hansch Analysis
log (1/dose) = k+k’(logP)-k”(logP)^2 + other terms
Training
determines parameters (P1, P2, etc) of math models using a training set
Testing
checks validity of model using a test set
Training and Testing
-both sets contain drugs with known activity data
-can be used to predict activity of drug without known activity data
Mathematical Model example
Activity = P1 * pi value
+ P2 * sig val
+ P3 * molecular weight
+P4 * LogP
+C
Machine learning
-provide computer with extensive set of data and let it determine QSAR itself
-no math model needed
-bases of SAR is not apparent
Machine learning Mech
Structures –(molecule vectorization)–> inputs –(multiple hidden layers)—> hidden layers —–>outputs (probabiility affinity)
Drug Discovery and development
Research –> target -(lead discovery)-> lead -(lead optimization)-> drug candidate –> trials
Lead discovery
-natural products
-antimetabolites
-structure-based drug design
-high throughput screen
-in silico drug screen
Lead optimization
-structure based drug design
-QSAR
-isoteric replacement
-prodrug
Antimetabolites
-analogues of endogenous metabolites
-resemble essential metabolite and competes with metabolite in physiological reactions
-folic acid vs methotrexate
Structure-based drug design
-based on 3D structure of target protein
-rational design vs in silico screen
