CSNAP (Chemical Similarity Network Analysis Pull-down) Web

Frequently Asked Questions:

What is the difference between CSNAP and ligand-based approaches or system-based drug-target networks (DTN)?

Although CSNAP utilize chemical similarity comparison similar to ligand-based approach, the ligand similarity only influences the overall network configuration of how compound ordered into the network while the target inference process is based on the generated network topology. Specifically, a conensus statistics of drug targets is computed by the target annotation frequency shared among neighbors of the query compounds within the networks. In contrast to conventional ligand-by-ligand analysis, many "emergent" properties of a compound set including chemotypes and target consensus can be revealed by this novel framework, making CSNAP particularly suitable for large-scale drug target deconvolution. CSNAP is also different from DTN approach which infers drug targets by bipartite networks of drug-target association patterns not involving chemical similarity relationship between ligands.

What validations have been conducted for the CSNAP approach?

We have validated CSNAP using a pseudo-library consisting of 216 compounds from 6 major drug classes retrieved from the Directory of Useful Decoy (DUD) sets and also applied CSNAP to identify drug targets for mitotic compounds identified in chemical screens. Please refer to the "About" section as well as the references for additional details.

What are the CSNAP Web input formats and how do I construct query molecules in these formats?

CSNAP Web accepts user input ligands in either SMILES or SDF formats, which are two of the most commonly used formats for molecular modeling and drug design. The SMILES format represents each molecule by a string of elementary symbols while the SDF format requires the spatial coordinate of the compounds and is more suitable for batch processing of large compound sets. To faciliate the querying process, we have also included the popular JME molecule editor in our web page, which allows users to construct any molecule on the canvas to be directly submitted as SMILES string on-site.

What fingerprint should I use for chemcial similarity searches?

CSNAP Web offers 4 different OBABEL fingerprints for chemical similarity searches including FP2, FP3, FP4 and MACCS, which represent different ways of converting a molecule into a binary string of 0 and 1 for structural comparison. FP2 fingerprint applies a path-based algorithm and is therefore more specific than the other fingerprints that are dependent on pre-defined substructure patterns. Thus, FP2 fingerprint can be used for any initial chemical searches while other fingerprints can be applied to broaden the hits if necessary.

What is the difference between Tanimoto coefficient (Tc) cutoff and Z-score cutoff? What cutoff values should I use?

Tanimoto coefficient (Tc) cutoff is based on the absolute chemical similarity while the Z-score cutoff is based on the relative similarity of the top 100 most similar compounds. In general, increasing similarity cutoff will increase the structure similarity of the retrieved compounds and thus prediction specificity. However, it is possible to retrieve additional compounds using lower cutoff if the original cutoff does not yield any hits. We recommend using the default option: Tc=1 and Z-score=2.5 for initial searches which works well for most compound classes.

Is it possible to identify targets for drugs that are potentially interacting with non-protein targets?

Yes, in the CSNAP Web search page, you can set the ChEMBL confidence score to "3" which will cover potential non-protein targets.

Is it possible to identify direct/functional targets for my compound query?

Yes, in the CSNAP search page, you can set the assay type "B" for biochemical (direct) targets or "F" for functional targets.

How do I intepret the CSNAP results?

The CSNAP Web output pages consists of five panels: chemical similarity network, ligand-target interaction fingerprint (LTIF), nodes and edges information and QuickGo search of ChEMBL target annotations. The query compounds are labeled in reds while the reference compounds are labeled in gray. The nodes and edge can be clicked to show addtional information in the bottom information panels. For a set of chemical analogs, the LTIF can be used to determine their target specificities/promiscuity. Please refer to "tutorial" and "help" page for more details.

Why did CSNAP not identify any targets for my query compounds?

It is possible that the submitted compounds have not been de-orphanized or may represent novel drug classes. Consequently, the target annotations of the compounds and/or its analogs are not currently in the ChEMBL database. To confirm this, the users can lower the Tc cutoff or apply different fingerprints to see if additional analogs can be identified.

Why the CSNAP webpage does not being displayed properly?

To ensure compatibility with the CSNAP Web, please update your web browsers to the latest versions. CSNAP Web has been tested and optimized for Firefox (version 30.0), Chrome (Version 36.0.1985.125) and Opera (vesion 12.16). Limited compatibility is observed for Internet Explorer. Some known issues including compound structure scaling problem in the output page. However, the structures can be restored by zooming using the browser tools. All other funtionalities are not affected by this operation.