Docking methods are now widely used and applied in different areas of research, although their performance may be questioned in challenging cases (see CAPRI results and reports). There is definitely a need for web services in this research field where the more user-friendly methods (such as autodock, autodock-vina, ...) require some installation before to use the software and some training and practice to be able to use it in the proper and more efficient way. Besides, the interaction between developers and users (specialists or non-specialists) is always stimulating and source of improvements and innovation. The paper should be published taking into account the recommendations and required clarifications.

From the scientific point of view, the predictive value of the EADock DSS engine is ascertained from a previous published work on a pretty large data set (Grosdidier et al., 2009). The web interface also includes (S3DB tab) a local database of manually curated complexes. Other clients are developed via SOAP using different coding schemes (PHP, python, perl). From the technical point of view, the interface is simple and well-designed and fulfills all the expectations of a user-friendly interface; the compatibility and robustness of the web service was checked and passed using three different web browsers. Visualizations of the docking results are proposed directly from the web (Jmol) or through an external visualization interface (Chimera) using the output data generated by the web server in the appropriate format. A forum of discussions is also included.

The web server was tested using two protein targets: one X-ray structure (PDB ID: 3NU3) and one NMR structure (PDB ID: 1BVE) of the HIV-1 protease both complexed with a high-affinity ligand: amprenavir (ZINC ID: 3809192) and DMP323 (ZINC ID: 18249417). "Very fast" and "accurate" types of docking were run on both targets using a more or less extended flexibility (3 or 5Å). The URL corresponding to those tests are the following: test 1: http://www.swissdock.ch/retrieve.php?job=853569854/HIV-1_protease_amprenavir.zip test 2: http://www.swissdock.ch/retrieve.php?job=1051415607/HIV-1_protease_amprenavir_accurate.zip test 3: http://www.swissdock.ch/retrieve.php?job=1059771564/3NU3_fast.zip test 4: http://www.swissdock.ch/retrieve.php?job=1825831932/1BVE_DMP323_fast.zip From those tests, a few recommendations are suggested: - include the references of the molecules ID and parameters used in the docking run, they should appear both on the web interface and in the zip archive to keep track of the molecules and parameters used; - include the definitions of the criteria used to rank the different binding modes (cluster, fullfitness and RMSD reference) as it is specified in the submit docking window.
The results obtained on tests 1 and 3 (using the same docking parameters) show some apparent discrepancies while the only difference comes from the way the protein target was selected: PDB code or PDB file upload in tests 1 and 3, respectively. The authors should clarify the reasons of those discrepancies.

P.S.: typos should be corrected, for example: page 3/line 8, page 3/line 17, ...