Learning for Protein Structure and Function Prediction

Machine Learning for Protein Structure and Function Prediction 

Machine learning is an established tool in many problem domains ranging from computer vision to stock markets to computational chemistry. A machine learning algorithm automatically discovers patterns in historical data to improve future 

Annual Reports in Computational Chemistry, Vol. 4 ISSN 1574-1400, DOI 10.1016/Sl574-1400(08)00003-0 

The outline of this chapter is as follows. The second section introduces a number of important supervised learning problems and illustrates how a biomolec-ular application can be cast in each problem formulation. Specifically, modeling protein-DNA interactions serves as the example for each of these formulations. The third section summarizes recent applications of machine learning to biomolec-ular modeling. The final section discusses current trends and future directions of machine learning applications to biomolecular modeling. 

---

For Kosan and others in the pharmaceutical industry, the intent is to learn enough about these enzymes from a structure-function viewpoint so they can be manipulated. Because polyketides are built by starting from one point and continuing sequentially along the pathway dictated by protein structure, a biochemist can trace through a molecule and make structure-function predictions for the assembly enzymes with reasonable accuracy. Using recombinant DNA methods,... 

Conventionally, the first attribute known about an enzyme used to be its function, usually in a crude extract. This property was screened for in microbial cultures or in tissue samples. The crude extract was then purified to homogeneity and the protein subjected to biochemical studies to learn of its pH and T profiles, its pi and subunit composition, catalytically important residues, and other properties. Proteolytic digestion of the protein with subsequent Edman degradation led to the primary sequence, but no information on the secondary structures such as a-heli-ces and [5-sheets or the folding in three dimensions of the polypeptide chain. The primary sequence could have been used to deduct the gene sequence but, with the degeneration of the code, several possibilities for certain amino acids occur, which makes prediction of the gene sequence a risk. 

---