Kihara Bioinformatics Lab Research Summary 2016

1

Daisuke KiharaProfessor of Biological Sciences, Computer Science

Bioinformatics Virtual drug screening Protein structure prediction Structure modeling for electron microscopy data

http://[email protected]:@kiharalab

PL-PatchSurfer2: Molecular Surface-Based Virtual Screening Method

2

Compound LibraryBinding Pocket

Adenosine-binding

(Shin,Christoffer, Wang, Kihara, JCIM, 2016)

Methods Structure EF1% BEDROC

PL-PSurfer2 X-ray 12.86 0.31

TBM 11.76 0.31

Autodock Vina

X-ray 8.63 0.33

TBM 1.68 0.09

DOCK6 X-ray 11.70 0.26

TBM 2.58 0.12

http://kiharalab.org

Benchmark on the DUD dataset

• Performs better than existing methods particularly on computationally modelled target protein structures

Protein Docking

Drug Virtual Screening

Electron Microscopy

Structure Data

Ligand Binding Pocket

Comparison

http://kiharalab.org 3

Protein Function Prediction

4

http://kiharalab.org/pfp_esg.php

http://kiharalab.org

Genome-Scale Detection of Moonlighting Proteins

•Proteins that have two distinct functions in a single chain

•Genome-scale prediction of moonlighting proteins

• 2.7 % of C. elegans, 7.8% of human, and 11.0 % of yeast proteins in their genomes are predicted to be moonlighting proteins

5

(Khan & Kihara, Bioinformatics 2016)

http://kiharalab.org

FoldsTksfE

TksfE=)s|p(f

kf

Bjk

Bjiji /|exp(/|exp(

Protein Folding Channel

Consider information theoretic noisy channel that transmit sequence (S) information to protein structures (F, folds)

6

P(fi|sj)

Boltzmann distribution:

(Magner, Szpankowski, Kihara, Scientific Reports, 2015)

http://kiharalab.org