Proteins are involved in almost all biological activities
and their structures and functions are determined by their one-dimensional
amino acids sequences. Since slow and expensive techniques by wet experiments,
the annotations of protein structure and function are far behind protein
sequencing, especially since the development of next generation sequencing
techniques. A practical way to solve this problem is to use computational
modeling methods. We developed a new template-based protein structure
prediction method SPARKS-X. The method improves the fold recognition by
employing probabilistic-based matching between predicted one-dimensional
structural properties of query sequence by SPINE-X and corresponding native
properties of template structures. This method has been consistently proven to
be one of the best methods on different benchmarks. This is further affirmed by
its top
rank in recent CASP blind tests on structure prediction. Based on SPARKS-X, we
have developed a SPOT-Seq package for predicting DNA-binding and RNA-binding
proteins by combining it with binding affinity prediction according to a knowledge-based
potential function DFIRE. This package enables a genome-wide scanning of
DNA-binding and RNA-binding proteins. Beyond making two-state binding or
non-binding prediction, the method can also provide a prediction of highly
accurate complex structure of protein-DNA and protein-RNA, respectively. These
details are extremely informative for experimental biologists.