ROSETTA scoring function p(sequence structure)=p(aa, aa,,aa,X) P(aa; aa, X c1,C2,…a X)≈∏Pa|) P(aa, XP(aa X) p(sequence structure) Pm P IP(aa, jE) E: reflects extent of burial P(aa,aa,,,r1) Dau is i P(aa Ei P(aa ei, r
ROSETTA scoring function P( sequence | structure) = P (aa1,aa 2 ,...aa | X) n P (aai ,aaj | X ) i< ∏ j P ( sequence | structure) ≈ Penv Ppair ∏ i P (aa1,aa 2 ,...aa | X ) ≈ P (aai | X ) n P (aai | X ) P (aa j | X ) Penv ∏ i = P (aai | Ei ) E reflects extent i of burial P (aai , aaj | Ei ,Ej ,rij ) P (aai | Ei ,rij ) P (aa j | Ej ,rij ) i< ∏ j Ppair =
RoSETTA Obstacles Enhancements Problem 1: generate lots of unrealistic decoys Filter based on contact order, quality of B-sheets, poor packing Problem 2: large search space Bias fragment picking by predicted secondary structure, faster computational algorithms Problem 3: low confidence in the result Fold many homologs of the target cluster the answers, report the cluster with highest occupancy
ROSETTA Obstacles & Enhancements • Problem 1: generate lots of unrealistic decoys – Filter based on contact order, quality of β-sheets, poor packing • Problem 2: large search space – Bias fragment picking by predicted secondary structure, faster computational algorithms • Problem 3: low confidence in the result – Fold many homologs of the target, cluster the answers, report the cluster with highest occupancy
ROSETTA performance at CASP4 was very impressive 17 /21 predictions had >50 residue fragments with rmsd <6.58 Occasionally found structures better than the best representative in the pdb(i.e. better than best-possible fold recognition performance)
ROSETTA performance at CASP4 was very impressive • 17/21 predictions had > 50 residue fragments with rmsd < 6.5Å • Occasionally found structures better than the best representative in the pdb (i.e. better than best-possible fold recognition performance)
6.4 A rmsd 5 Cys pairs correct new folds 4.9A rmsd Bonneau, R,J Tsai, I Ruczinski, D Chivian, C Rohl, CE Strauss, and D Baker. " Rosetta in CASP4: Progress in ab Initio Protein Structure Prediction. Proteins Suppl 5 (2001): 119-26
new folds 4.9 Å rmsd 6.4 Å rmsd 5 Cys pairs correct Bonneau, R, J Tsai, I Ruczinski, D Chivian, C Rohl, CE Strauss, and D Baker. "Rosetta in CASP4: Progress in Ab Initio Protein Structure Prediction." Proteins Suppl 5 (2001): 119-26
Flowchart for rosetta as used in CASP5 Bradley, P, D Chivian, J Meiler, KM Misura, CA Rohl, WR Schief, WJ Wedemeyer, O Schueler- Furman, P Murphy J Schonbrun, CE Strauss, and D Baker. "Rosetta Predictions in CASP5: Successes, Failures, and Prospects for Complete Automation. "Proteins 53, Suppl 6(2003 ): 457-68
Flowchart for ROSETTA as used in CASP5 Bradley, P, D Chivian, J Meiler, KM Misura, CA Rohl, WR Schief, WJ Wedemeyer, O Schueler-Furman, P Murphy, J Schonbrun, CE Strauss, and D Baker. "Rosetta Predictions in CASP5: Successes, Failures, and Prospects for Complete Automation." Proteins 53, Suppl 6 (2003): 457-68