Best practices for chemical data curation and QSAR model ...

More documents

Recommendations

Info

Why models may fail • Incorrect data (structures and activities) in the dataset • Modeling set is too small • No external validation • Incorrect selection of an external test set • Incorrect division of a dataset into training and test sets • Incorrect measure of prediction accuracy • Insufficient statistical criteria to estimate predictive power of models • Lack or incorrect definition of applicability domain • No Y-randomization test (overfitness) • Presence of leverage (structure) and activity outliers Also, see Dearden JC, Cronin MT, Kaiser KL. How not to develop a quantitative structure-activity or structure-property relationship (QSAR/QSPR). SAR QSAR Environ Res. 2009;20(3-4):241-66
Some reasons why QSAR models may fail: using incorrect target function in classification QSAR for biased datasets: • A typical target function (Classification Rate): CR=N(classified correctly)/N(total) A dataset: Class 1: 80 compounds Class 2: 20 compounds Model: assign all compounds to Class 1. Target function: CR=0.8 The model appears to have high classification accuracy • Better target function: CCR=0.5(Sensitivity+Specificity) In the above example, CCR = 0.5 • General formula: CCR = 1 K K ∑ k = 1 N N corr k total k K – the number of classes QSAR Pill N k corr – the number of compounds of class k assigned to class k N k total – total number of compounds of class k • For categorical response variable, target functions can depend also on the absolute errors (differences between predicted and observed classes).
Page 1 and 2: Alexander Tropsha Laboratory for Mo
Page 3 and 4: The rumors of QSAR demise have been
Page 5 and 6: Quantitative Structure Property Rel
Page 7 and 8: QSAR Modeling appears easy… Goal:
Page 9 and 10: BEWARE OF q 2 (Kubinyi paradox)!!!
Page 11: Complexity of QSAR modeling: Choice
Page 15 and 16: Some reasons why QSAR models may fa
Page 17 and 18: Applicability domain of QSAR models
Page 23 and 24: Recently, D.Young et al. pointed ou
Page 25 and 26: Revising QSAR Modeling Process : Pr
Page 27 and 28: The importance of data curation: Wh
Page 29 and 30: Division of the Dataset into Three
Page 31 and 32: The OECD Principles of model valida
Page 33 and 34: Database Mining Reveals Unique Chem
Page 35 and 36: Recent examples of experimentally v
Page 37 and 38: Study Design (AmpC β-lactamase dat
Page 39 and 40: Descriptor Interpretation Rank Desc
Page 41 and 42: One “inactive” compound (CID 69
Page 43 and 44: Target-specific filter construction
Page 45 and 46: Docking enrichment plot for DHFR us
Page 47 and 48: Target related ligands Disease-Targ
Page 49 and 50: Comparison of the QSAR Approaches t
Page 51 and 52: Querying the cmap with Alzheimer’
Page 53 and 54: Selected Common Hits from QSAR and
Page 55 and 56: Raloxifene is predicted to bind sev
Page 57 and 58: Importance of hypothesis fusion WDI
Page 59 and 60: Chemocentric view of biological dat
Page 61 and 62: Poor global relationships between i
Page 63 and 64:
Use of hybrid descriptors for struc
Page 65 and 66:
A new “hierarchical QSAR” appro
Page 67 and 68:
Case Study 1 Recently 1 , 51 divers
Page 69 and 70:
CS1. QNTR modeling results of 44 di
Page 71 and 72:
International Virtual Collaboratory
Page 73 and 74:
Overview of the Approaches (15 meth
Page 75 and 76:
Which model is best • Observation
Page 77 and 78:
Principles of “Safe” QSAR model
Page 79:
Final Word Nothing that worth knowi
show all

Best practices for chemical data curation and QSAR model ...

You also want an ePaper? Increase the reach of your titles

Delete template?

Save as template?