(661i) Loop Structure Prediction for Fixed-Stem Geometries In Proteins

Loop structure prediction is an important intermediate step towards the tertiary structure prediction of proteins[1,2,3,4]. In the absence of an inherent local hydrogen bonding pattern, loop structure prediction can be considered a mini ab initio structure prediction problem. Here, we present an algorithm for the prediction of loop structures in fixed-stem systems. In fixed-stem geometries, the knowledge of the coordinates of the secondary structure elements bordering a loop is known, while the structure of the loop itself is unknown. The problem forms a critical intermediate step in the elucidation of protein structures using database driven methods like homology modeling and fold recognition.

 Using a distribution of loop phi/psi angles generated from a large, non-redundant database, initial structures for the loop between two given secondary structure elements are generated. While initial bounds on the dihedral angles in the loop region are generic, distance and dihedral angle constraints derived from the knowledge of the structure of stem residues provide restrictions on the possible prediction of loop structures. Three fast side chain rotamer optimization steps are utilized to alleviate steric clashes in the randomly generated initial structures, thus providing better starting points for structure optimization. The structures are now subjected to energy minimization, using the all atom force field ECEPP/3. The problem is formulated as a constrained non-linear optimization problem; with the constraints representing distance and dihedral angle bounds on the loop and stem residues. At the end of the optimization step, the generated conformers are clustered using a traveling salesman problem based clustering algorithm, ICON [5]. ICON is an iterative algorithm which treats each conformer generated as a node in a TSP, and finds the best path to connect these nodes. An integer optimization (ILP) based model is implemented to rigorously find cluster boundaries, which lead to the elimination of sparse clusters at each iteration. The densest clusters are then collected, and improved dihedral angle bounds on the backbone angles of each of the loop residues are generated for the next iteration of the algorithm. At each iteration, new conformers are tested against previously generated conformers to eliminate duplication of initial structures. The algorithm is being tested on a large set of loops between the lengths of 5 and 20. The initial results and comparisons with experimentally elucidated structures show a high degree of accuracy in predicting tight bounds on the backbone angles of loop regions, thus displaying the method’s efficacy in being used as the loop generation algorithm for database driven structure prediction algorithms.

Bibliography

[1] McAllister SR and Floudas CA (2010) An improved hybrid global optimization method for protein tertiary structure prediction, Comput. Optim. Appl. 45, 377-413

[2] Floudas CA, Fung HK, McAllister SR, Monnigmann M and Rajgaria R (2006) Advances in Protein Structure Prediction and De Novo Protein Design: A Review, Chem Engg. Sci., 61, 966-988

[3] Floudas CA (2007) Computational methods in protein structure prediction, Biotech. BioEng., 97, 207-213

[4] Subramani A, Wei Y and Floudas CA (2011) ASTRO-FOLD 2.0: An enhanced framework for protein structure prediction, submitted.

[5] Subramani A, DiMaggio PA and Floudas CA (2009) Selecting high quality Protein structures from Diverse Conformational Ensembles, Biophys J, 97, 1728-1736