photo Sign

Carlos J. Camacho
Structural Bioinformatics Lab
Department of Computational Biology
University of Pittsburgh
empty7
border
spaceleft

Research Projects

Protein-protein interactions

Protein binding dynamics

Protein docking

Protein-DNA interactions

Homology modeling
CASP5 results

Genomics and molecular evolution

Thermodynamic & dynamics of protein folding







Sign



Homology Modeling and CASP5 performance:

Homology Model Server
The server, developed by Grad.Student Jahnavi Prasad and myself, can be reached directly at Consensus [1], [2]. The main feature of the server is to identify regions of high structural similarity (HSS) between a target sequence and a template. If no template is given, the server will identify one for comparative model targets.
The predicted HSS regions are typically less than 2.5 A RMSD from the crystal structure, as estimated from the recent blind experiment at the CASP5 meeting. The output of the server also includes an optimal alignment between target and template. In a few weeks, the server will be upgraded to output full 3-D PDB models (instead of just the alignment).
Using the output of our server, we have also validated a (manual) Comparative Modeling technique to predict accurate homology models. To validate our methods, our performance on the recent CASP5 experiment (where we submitted models for all targets) is summarized below. We note that 215 research groups participated in CASP5, submitting a total of 22909 models).

CASP5 (Asilomar, 2002):

CASP5/CAFASP3 results
Number of groups and servers contributing: 215
Number of models designated as Nr. 1: 7627

Summary of the performance of our group on the Comparative Modeling category at CAFASP3 (servers) and CASP5 (manual), as well as in the fold recognition (FR) and new fold (NF) category is listed. Analysis includes ONLY models submitted as Nr. 1 prediction.

GROUP______________MEAN LENGTH________MEAN RMSD__NOTE
NAME_____CODE___TARGET__PREDICTED____(per Residue)
--------------------------------------------------------------------------------
CAFASP3__completely automatic structure prediction for CM targets.

Camacho__(#098)____158_______89__________2.4 A_______(1)

CASP5__manual (Top CM Predictors based on RMSD and LENGTH)

Camacho__(#099)____161______139__________4.8 A_______(2)

Bujnicki___(#020)____161______156__________5.2 A_______(2)

Fischer____(#427)____161______149__________5.7 A_______(2)

Baker_____(#002)____161______149__________5.9 A_______(2)

CASP5__manual (FR & NF category based on RMSD and Alignment LENGTH)

Camacho__(#099)____120________68_________11.7 A______(3)

Fischer____(#427)____120_______118_________15.2 A______(3)

Baker_____(#002)____120_______110_________13.0 A______(3)
--------------------------------------------------------------------
(1) Average is over the 44 comparative model (CM) targets of CASP5 that our server was able to automatically find a template in the PDB. A total of 52 servers participated in this blind experiment, all other servers predicted models with more than twice our mean RMSD. (See, also, http://www.pdg.cnb.uam.es/CAFASP3Contacts_RESULTS/index.html)

(2) Average is over all 51 CM targets.

(3) Average is over all 30 FR & NF targets.
-------------------------------------------------------------------


Publications | Curriculum Vitae | Group/Collaborators | Useful Links
CONTACT

Carlos J. Camacho
Associate Professor
Department of Computational Biology
Department of Molecular Genetics and Biochemistry
University of Pittsburgh
W1041 Biomedical Science Tower
200 Lothrop St.
Pittsburgh, PA 15261
Office: (412) 648-7776
Fax: (412) 648-3163
E-mail: ccamacho@pitt.edu














spaceleft

Public Servers

Protein interactions:
ClusPro
NEW Validation CAPRI3-5.PDF

Homology modeling:
Consensus
CAFASP3 benchmark
NEW Modeling side chains.PDF

Protein-protein interactions:
NEW FastContact

Splitting structural domain:
DomainSplit


Software

NEW FastContact PDF


Public Databases

Docking decoys