Bioinformatics Methods and Applications Dr. Hongyu Zhang Ceres Inc.

Bioinformatics

Methods and Applications

Dr. Hongyu Zhang

Ceres Inc.

Goals of the talk

• The major battle fields in Bioinformatics research

• The most popular weapons used in the battle

History

• Human genome project

• Overlapping with other branches– Computational Biology– Biocomputing– Biostatistics– Cheminfomatics

The Central Dogma ofMolecular Biology

DNA RNA ProteinTranscription Translation

Major battle fields in bioinformatics

• DNA– Genome sequencing– Gene discovery

• mRNA– Micro-array analysis– Sequencing

• Protein– Structure modeling and prediction– Proteomics

• …

Major weapons• Computational algorithm

– Hash method– Dynamic algorithm– String and Tree (binary, suffix)– Clustering

• Probability and Statistical theory and methods– Bayesian theorem, Markov chain (HMM), Principle component– Monte Carlo simulation– Neural Network

• Physical chemistry– Functions to describe the physical chemistry interactions in bio-molecules– Molecular mechanics, Molecular dynamics algorithm

• Data storage and access– Database: Oracle, MySQL etc.– Web interface

• Large-scale computing platform– Hardware– Software

Genome sequencing: Celera shotgun assemblyVenter et al. 2001

Gene discoverybased on sequence comparison

• Finding new genes based on their sequence similarity and evolution relationship with known genes

• Methods– Hash-based database search method, like BLAST

(PSI-BLAST), FASTA, BLAT etc.– Sequence alignment using Dynamic Programming

algorithm

BLAST database search (http://www.ncbi.nih.gov/BLAST/)

Query sequence

Database sequences

Querydatabase

Sequence alignment

BLAST||| |BLA-T

• Example

• Programs

• CLUSTALW • DIALIGN

Dynamics algorithm

,,1

,1,

,1,1

, max

Aiji

Biji

BjAiji

ji

SH

SH

SH

H

Sequence A = (A1, A2, …, Ai, ..., Am)

Sequence B = (B1, B2, …, Bj, …, An)

Ab initio gene prediction methods

• Statistics based gene prediction– Nucleotides distribution frequencies in the cod

ing regions – Exon/Intron boundary signal

• Examples– GenScan, Burge and Karlin 1997 – Fgenesh, Solovyev and Salamov 1994

Hybrid gene prediction method

• Example: Celera Otto program– BLAST against Refseq database– BLAST against EST database, other genomic

sequences etc.– Genscan, Fgenesh

Problems in Gene discovery

• Example: Given a cDNA sequence, find its true location in the genome map among lots of alternatives

1

1’ 2’ 3’

2 3

Genomic component

Query transcript/protein

Two-step solution

1. BLAST search of the cDNA sequence against the whole genome map

2. Using an LIS algorithm to find the correct genomic component hit

Cutoffesifhspll

hspl

jiijij

i 0},,max{

}{

0

00

inilLIS

0max

Phylogenetic analysis

• Goal: study the function and evolution relationship among a group of genes– Divide homologous genes into function families – Find the evolution relationship between the ortholog g

enes belonging to different species (e.g., the theory of Out of Africa)

• Methods– Hierarchical Clustering– Neighbore-joining etc.

• PHYLIP program, Univ. of Washington

Micro-array analysis

• Expression-genomics

• Primary goals

– Look for the genes with different expression levels between experiments, which are candidates of functional genes

– Look for the group of genes that have correlated gene expression levels, which could suggest that they are in the same biological pathway

• Methods– General probability and statistics methods– Dimension reduction

• Principle components• Lowess

– Clustering

• Tools– S-plus, R

Example

• Herbicide– Plants was treated with herbicide to observe

the gene expression profiles in a series of time steps.

– The genes that appeared right before plant dies (12 hours) are the possible “death” genes

– If we knock down the “death” genes in the normal plants, they could last longer time than the herbs.

Protein structure prediction

• Why is protein structure important?– The functions of a gene depend on its translat

ed protein structure • Protein binding with its ligands• Protein-protein interactions

– A protein molecule usually keeps one stable structure under normal physiological conditions (Anfinson, 1960es)

– Drug design• Docking and high throughput drug screening.

Sequence

Protein structure

Function

Bioinformatics

Protein structure prediction methods

Protein sequence

Database search

Sequence alignment

Select template structure

Build conserved regions first

Loop modeling

Build side-chains

Optimizing

Homology modeling procedure

Homology modeling programs

• Academic software– MODELER, Sali A.– COMPOSER, Blundell T.– SWISS-MODEL – Rasmol (graphics)

• Commercial software– QUANTA, MSI inc.– SYBYL, TRIPOS inc.

Threading• Find the best fold candidates among a limited number of

choices• Add 3D information to the score function of dynamic prog

ramming

Ab initio protein structure principle

• Threading programs– Topits, Eisenberg D.– Threader, Jones D.– ProSup, Sipple M– 123D, Alexandra N.

• Ab initio programs– Rosetta, David Baker

Current status in the protein structure prediction field

• Moult J., CASP (Critical Assessment of Techniques for Protein Structure Prediction).

• Homology modeling is very mature already

• Threading and Ab initio method have been used in industry

• Structure genomics

Large scale computing platform

• Hardware– Super-computers

• Cray/SGI• DEC/Compaq• Intel

– Linux clusters– Blade

• Software– Parallel computing (MPP, P

VM etc.)– Linux – Grid computing: the Globus

Project

Linux clusters

Data storage and access

• Bioinformatics is producing huge amount of data each day– How to organize and store data – How to access data

• Database software– Commercial

• Oracle, DB2, Sybase

– Freeware• MySQL, PostgreSQL

Data store and access• Bioinformatics is producing huge amount of data each day

– How to organize and store data – How to access data

• Database software– Commercial

• Oracle, DB2, Sybase– Freeware

• MySQL, PostgreSQL

• Current popular database– DNA, protein sequence, like Genbank, SwisProt, PIR etc.– Protein structure, like PDB, Scop– DNA, mRNA, protein function, like GO, PFAM

Database example: Gene Ontology (GO)

Molecular function

Biologicalprocess

Cellularcomponent

Data access

• Web interface– Protocol

• CGI, JSP, ASP

– Computer languages• Perl, Java, C/C++, Visual Basic, Visual C++

Forth looking

• Where are the markets– Develop new programs– Assemble current programs to build more efficient data mining

pipelines– Data storage and access– Integrate the current database to use them more effectively– Computing platform, including hardware, software support,

consulting etc.

• What we can offer– Multi-talents– Team work– Networking

http://www.hongyu.org/paper/bioinformatics.ppt