Home
This Title All WIREs
WIREs RSS Feed
How to cite this WIREs title:
WIREs Comput Mol Sci
Impact Factor: 16.778

Modern 2D QSAR for drug discovery

Full article on Wiley Online Library:   HTML PDF

Can't access this content? Tell your librarian.

2D QSAR is a powerful tool for explaining the relationships between chemical structure and experimental observations. Key elements of the method are the numerical descriptors used to translate a chemical structure into mathematical variables, the quality of the observed data and the statistical methods used to derive the relationships between the observations and the descriptors. There are some caveats to what is essentially a simple procedure: overfitting of the data, domain applicability to new structures and making good error estimates for each prediction. 2D QSAR models are used routinely during the process of optimization of a chemical series towards a candidate for clinical trials. As more knowledge is gained in this area, 2D QSARs will become acceptable surrogates for experimental observations. WIREs Comput Mol Sci 2014, 4:505–522. doi: 10.1002/wcms.1187 This article is categorized under: Structure and Mechanism > Molecular Structures Computer and Information Science > Chemoinformatics
Assessing the predictive performance of a model with correlation charts and statistics.
[ Normal View | Magnified View ]
The effect of parameterization on prediction accuracy of the training and test set compounds. Based upon a figure by Hastie et al.
[ Normal View | Magnified View ]
The effect of increasing parameters on model fit. The generating model was a quadratic function with a random Gaussian‐distributed error term.
[ Normal View | Magnified View ]
A decision tree for predicting blood brain barrier permeation, as described by Suenderhauf et al. CNSp+ represents log PS measurement of > −2 and CNSp – represents a log PS measurement of < −3; alog P is the Ghose–Crippen predicted partition coefficient; fPSA3 is the fraction polar surface area; #RotBonds is the number of rotatable bonds; and #Acceptors is the number of hydrogen bond acceptors.
[ Normal View | Magnified View ]
The components of a 2D QSAR analysis.
[ Normal View | Magnified View ]
Receiver operating characteristic (ROC) curve generated from data obtained from Clark. log BB measurements and predictions were classified as CNS+ if log BB> − 0.5, and CNS– if log BB<−0.5. The AUC is 0.93.
[ Normal View | Magnified View ]
Confusion matrix for evaluating classification models generated from data obtained from Clark. The measurements and predictions from the DECII model were classified as BB+ if the log BB was greater than −0.5, and vice versa.
[ Normal View | Magnified View ]
Excessive leverage by a single isolated data point confounds model building.
[ Normal View | Magnified View ]
Anscombe's Quartet. For all four correlations: N = 11; STDEVy = 2.03; RMSE = 2.45; r2 = 0.67; and q2 = −0.45 or −0.46.
[ Normal View | Magnified View ]

Browse by Topic

Computer and Information Science > Chemoinformatics
Structure and Mechanism > Molecular Structures

Access to this WIREs title is by subscription only.

Recommend to Your
Librarian Now!

The latest WIREs articles in your inbox

Sign Up for Article Alerts