Outline Introduc4on to networks. Network alignment. 1 4/24/09 - - PDF document

4/24/09 1

CSCI1950‐Z Computa4onal Methods for Biology Lecture 21

Ben Raphael April 20, 2009

hGp://cs.brown.edu/courses/csci1950‐z/

Outline

• Introduc4on to networks.
• Network alignment.

4/24/09 2

Signaling Networks Networks and Cancer

Tumor Sequencing Project: Muta4ons in 188 lung cancer pa4ents. The Cancer Genome Atlas: Muta4ons in ~200 brain cancer pa4ents.

Are these results surprising?

4/24/09 3

A quick overview of signaling networks

• PDF slides.

Protein Interac4on Networks

• Proteins rarely func4on in isola4on, protein interac4ons

affect all processes in a cell.

• Forms of protein‐protein interac4ons:

– Modifica4on, complexa4on [Cardelli, 2005].

phosphoryla4on protein complex

4/24/09 4

Big Problem

LACK OF DATA! Pictures on previous slides summarize decades

• f experimental efforts.

High‐throughput DNA sequencing

Genomes Individual genes Interac4on Network Individual interac4ons

• r pathways.

????

How are protein‐protein interac4on networks derived?

Yeast two‐hybrid screens

4/24/09 5

How are protein‐protein interac4on networks derived?

Protein purifica4on and separa4on

Protein Interac4on Networks

(Over?)simplify interac4ons between proteins as a binary, sta4c rela4onship. Protein‐Protein Interac4on network – an undirected graph (usually)

• Nodes: protein
• Edges: interac4ons
• Edges may have weights

indica4ng confidence. – Yeast DIP network: ~5K proteins, ~18K interac4ons – Fly DIP network: ~7K proteins, ~20K interac4ons. – Human. ~20K protein. ~50K interac4ons.

PPI network

4/24/09 6

Computa4onal Problems

• 1. Comparing Networks Across Species
• 2. Classifying Network Topology

– Finding paths, cliques, dense subnetworks, etc.

• 3. Using networks to explain data

– Dependencies revealed by network topology

• 4. Modeling dynamics of networks

Alignment

Networks Evolve via gain/loss of proteins

• r interac4ons (?)

Sequences

Evolve via subs4tu4ons Conserva4on implies func4on EFTPPVQAAYQKVVAG DFNPNVQAAFQKVVAG mouse human

4/24/09 7

By similar intui4on, subnetworks conserved across species are likely func4onal modules

Mo4va4on Network Alignment

“Conserved” means two subgraphs contain proteins serving similar func4ons, having similar interac4on profiles

– Key word is similar, not iden4cal

mismatch/subs4tu4on

4/24/09 8

Alignment Analogy

Sharan and Ideker. Modeling cellular machinery through biological network

• comparison. Nature Biotechnology 24, pp. 427‐433, 2006

Earlier approaches: interologs

• Interac4ons conserved in orthologs

– Orthology (descended from common ancestor) is a fuzzy no4on – Sequence similarity not necessary for conserva4on of func4on

4/24/09 9

Complica4ons

• Protein sequence similarity not 1‐1

– Orthologs – Paralogs

• Interac4on data:

– Noisy – Incomplete – Dynamic

• Computa4onal tractability

Network Alignment

Sharan and Ideker. Modeling cellular machinery through biological network

• comparison. Nature Biotechnology 24, pp. 427‐433, 2006

4/24/09 10

The Network Alignment Problem

Given: k different interac4on networks belonging to different species, Find: Conserved sub‐networks within these networks Conserved defined by protein sequence similarity (node similarity) and interac4on similarity (network topology similarity)

• Goal: iden4fy conserved pathways (chains)
• Idea: can be done efficiently by dynamic

programming if networks are DAGs

Kelley et al (2003)

D D’

+ match

PathBLAST

C X’

+ mismatch

B

+ gap

A A ’

Score: match

4/24/09 11

Why paths? PathBLAST

(Kelley, et al. PNAS 2003)

• Find conserved

pathways in protein interac4on maps of two species

• Model & Scoring:

(Whiteboard)