Andrzej Kloczkowski Research Interests

Adjunct Professor
Computational molecular biology
Structural bioinformatics
Prediction of protein structure, function and dynamics
Biomolecular simulations

My research interests focus on various aspects of computational molecular biology and structural bioinformatics.

Lattice models of proteins are used to study the physical nature of protein folding. Simple two-dimensional and three-dimensional lattices, such as the square, the triangular or the cubic lattice, greatly simplify the conformational space for the study of folding, and in the search of the protein native state with the lowest energy. Protein conformations are modeled as compact self-avoiding walks on the lattice that are named in mathematics as: Hamiltonian paths, and Hamiltonian circuits (for circular chains without ends). We have developed a mathematical formalism called the transfer matrix method that allows us to replace the combinatorial geometric problem for the generation and enumeration of such compact conformations with a simple algebraic problem of matrix multiplication. We are also studying protein designability, which tells us why protein sequences are not random strings of amino acids but instead show regular patterns that encode protein structures. Reduced models of proteins based both on the lattice representation and the reduced binary hydrophobic-polar (HP) amino acid alphabet lead to highly interesting results that shed light on the evolutionary relationships among proteins.

Prediction of protein structures from the amino acid sequence is one of the most important problems in biology. Recent progress in the high-throughput sequencing projects have produced massive numbers of protein sequences, for which only the amino acid sequences are known, but whose crystallographic structures have not yet been determined. Additionally the gap between the number of experimentally determined protein structures, and the number of known sequences continues to accelerate. We predict both the secondary structures of proteins, and their tertiary structures using homology, fold recognition and ab initio methods. We have developed a publicly available web server http://gor.bb.iastate.edu/cdm/ for protein secondary structure prediction. We are collaborating with various research groups, including winners of the CASP competitions in the blind prediction of protein structure from the amino acid sequence, to develop a methodology of obtaining high resolution protein models from lower resolution data.

Protein dynamics. We use elastic network models to predict large-scale functional motions of biological molecules that are crucial for their function. This approach, which originates from our previous rubber-like elasticity studies of random polymer networks, has been extremely successful. It allows us to generate pathways for protein conformational transitions between ‘open’ and ‘closed’ forms and other transitions such as domain swapping. It enables us to predict the order of breaking contacts upon mechanical unfolding in single molecule pulling experiments. Our studies suggest that structure controls the global motions of proteins not only in their native states, but also in the transient conformations. Our results replace the prevalent static view of proteins with a more dynamic one, and show that flexibility is essential for protein biological function.

Prediction of protein functions. The functional annotation of proteins obtained from mass-scale genome sequencing is a current major problem. Sequence-based annotations of proteins that have sequence identities below 40% in comparison with their annotated homologs are unreliable. We use both structure prediction and resulting protein dynamics computations to annotate such proteins. We are developing new methodologies to perform such annotations on the genome scale. We are especially interested in agriculturally important genomes, such as bovine or maize.

Protein packing and statistical potentials. We study packing of residues in proteins, orientational distribution of side chains within clusters of neighboring residues. We compare these distributions with directional vectors in simple 3D lattices, and regular and semi-regular polyhedra.  We develop many-body statistical contact potentials for proteins, based on such polyhedral models, which are superior for identifying native structure among decoys in threading.

Prediction of binding sites, prediction of phosphorylation sites and other post-translationally modified sites. We use statistical machine learning methods such as support vector machines (SVM) to predict binding sites, phosphorylation sites, and other residue specific protein properties, and develop web servers for such predictions.

Systems Biology. We integrate a variety of genomic, proteomic, metabolomic, and interactome data by using network models, aiming at the development of models of cells.