Surflex-Sim

Molecular Alignment and Virtual Screening

Overview 

The majority of drug targets for small molecule therapeutics are proteins whose three-dimensional structure is not known to sufficient resolution to permit structure-based design. Complementary three-dimensional ligand-based design approaches require some hypothesis of bioactive ligand conformation and alignment.

Optimal superposition of nicotine and a competitive oxazole derivative by morphological similarity. The translucent grey surface represents the overall molecular volume of the aligned molecules. Differences in measurements made of nicotine and the oxazole derivative from the observation points in space are shown. The red surface indicates differences in the hydrophobic surfaces being observed, the blue and green surface indicates the high degree of electrostatic potential overlap between the two molecules. The major difference arises from the protruding methyl on the oxazole, not the difference in 2D substructure of the oxazole and pyridine.

Surflex-Sim rapidly and automatically optimizes the alignment of molecules maximizing their three-dimensional similarity. Surflex-Sim uses a surface-based morphological similarity function while minimizing the overall molecular volume of the aligned structures. This approach has been demonstrated to outperform other techniques in comparisons of correlation to biological activity.

The morphological similarity approach of Surflex-Sim effectively facilitates scaffold hopping, thus allowing the user to find alternative chemical scaffolds without the same toxicity and potential intellectual property issues as the given lead. This leads to truly novel lead series with distinct properties while preserving those structural properties that are essential for biological activity.

Surflex-Sim is now fully integrated into Tripos' market-leading SYBYL product and available for immediate release on Tripos' Download site.

Surflex-Sim Brochure (122k)

Key Features

• Pairwise Alignments and Morphological Similarity: Surflex-Sim rapidly optimizes the pose of a query to maximize three-dimensional similarity to an object molecule. No input for mapping or guidance is necessary. Molecules judged to be similar tend to bind to the same proteins.

• Optimal Ensemble Alignment: Surflex-Sim produces an optimal superposition for a small number of competitive ligands, maximizing pairwise similarities while minimizing total volume.

• Drug and Target Similarities: It is possible to rationalize off-target effects of drugs and to predict them. Drug pairs sharing a target have significantly higher similarity than drug pairs sharing no target, and target pairs with no overlap in annotated drug specificity share lower similarity than pairs with increasing overlap.

• Ligand-Based Virtual Screening: Superpositioning multiple molecules forms a hypothesis as to their preferred binding mode to a target. Surflex-Sim functions as a similarity program to rank a set of input molecules according to their degree of fit to the hypothesis. Models generated with Surflex-Sim have the ability to identify cognate drugs from a background of other drug molecules, producing fewer false positives and more true hits.

Validation

Surflex-Sim has been extensively validated on 59 different datasets across a wide variety of targets.

The Receiver Operated Characteristic (ROC) Area Under the Curve (AUC) is a threshold independent score for the fractions of true positives and false negatives in compound screening. A ROC AUC of 1 represents a 100% true positive rate at a false positive rate of 0%. A ROC AUC of 0.5 represents a true positive and a false positive rate which is equivalent to a random selection.
The displayed histogram summarizes the ROC scores for 59 arbitrary targets. The datasets used in the study are available for free download from http://www.biopharmics.com/downloads.html.