Module 3 Journal Club

Koonin et al

• The Net of life
• microbial phylogenetic network
• with HGT observe patchy distribution of orthologs within clades
• can detect gene loss events as well

property of hgt network

• scale-free - frequence versus number of nodes - node connectivity displays power law distribution - the rich get righer and the poor get poorer
• HGT champions - "hubs" small number of species each linked to 30-50
• can't really determine transfer direction, can infer through parsimony
• only dealt with extant species

Transcriptional Regulatory Networks in Saccharomyces Cerevisiae

• put all pathways to one figure
• talk given by Yevgeny

1. Why this paper
2. experimental design

• transcription - protein that regulates transcription of genes
• tag the transcription factors
• Chromatin IP to enrich promoters bound to regulator in vivo
• microarray to hybridize - just to identify promoters bound to regulator in vivo ... don't need to do this anymore, this is old technology
• "next-generation sequencing"

results

• 1 transcription factor can bind to more than 1 promoter region - makes more robust

elementary units

1. autoregulation
2. multi-component loop
3. feedforward loop
4. single-input loop - bunch of
5. multi-input loop
6. regulator chain - provides temporal control

build network

• tie pathways to cell cycle
• automatically assembled

network motifs: simple building blocks of networks

• r. milo et al science 2002

Barabasi et al

• nature paper is translational
• human interactome - noisy and incomplete
  • transcriptional regulatory network
  • virus host network
  • metabolic network
  • etc
• properties of disease network - are desease gene a hub?
  • disease causing genes has more connection - but sample bias?
  • certain genes are mentioned in proposal simply to get grant funding
• hub genes important in all tissues, but disease genes are more tissue specific
• car mechanic analogy - no map of human body system to fix

network based analysis of breast cancer

• distance metastases are the ones that kill you
• every gene chips
• if you look at one gene, not gonna get right answer, got to look at modules
• goal: develop a kit to predict whether patient is susceptible to metastasis
• network based clasification achieves higher accuracy in prediction
• disease = subnetworks, pathway, process is messed up.
  • there's hope for GWAS
• stop metastasis - you have a fighting chance

schadt et al eric

• which SNPs are causal and which are just along for the ride?
• new paradigm - genetic perturbation has molecular effect which causes disease, not direct
• 20 snps, don't know which one's causal, use instead for hypothesis generation, which protein is affected by the snp
• eqtl - expressed quantitative trait locus
• network approach
  • genetic and environmental perturbations affect network
• disease diagnosis based on
• multiple testing correction - a big problem
• complex traits , within a pedigree you might see significant snp but across pop you won't
• bottom line is one disease manifestation may actually be 50,000 diseases
• germ line mutation versus somatic cell mutations