Seminar on Computational Learning and
Adaptation
Data Mining for Understanding Flavors of Protein Disorder
Zoran Obradovic
Center for Information Science and Technology
Temple University
Philadelphia, PA
zoran@divac.ist.temple.edu
www.ist.temple.edu/~zoran
Regions or in some cases whole proteins could lack a fixed structure in the native state. Such intrinsic disorder is often utilized for protein function. Our previous studies provided strong evidence that: (1) disorder is a very common element of protein structure; (2) the strength of disorder prediction is correlated with sequence complexity; and (3) eukaryotes evidently have a much larger fraction of proteins with intrinsic disorder than eubacteria or archaebacteria. This talk will focus on detecting and characterizing different types of protein disorder from primary sequence information. We will describe a supervised clustering algorithm that employs competing predictors to partition a set of disordered proteins into subsets with similar properties. An extensive statistical and qualitative evaluation of the resulting partitions provide strong evidence that: (i) there are at least 3 distinct flavors of protein disorders; (ii) all 3 flavors are common in nature, but the flavors statistics are very different among 32 genomes; (iii) different disorder flavors are dominant in bacteria, archaea and eucaryotes.
Date: Tuesday, January 22
|
Time: 4:15-5:30PM
|
Place: Ventura 17
|
Return to the seminar schedule