Sep 20, 2011

Beyond the Genome 2011: Day 1: Cancer Genomics

Today's presentations highlighted how challenging it is to study the cancer genome. Also highlighted the need for bioinformaticians to get a grip on human genetics in order to correctly interpret genomic datasets, specially sequencing results.



Elaine Mardis (WashU)
A glimpse at tumor genome evolution


Sam Aparicio (BC cancer agency, Canada)
Characterizing triple-negative breast cancer tumors


Charles Perou (UNC Chapel Hill)
Comparative genomic analysis of mouse and human mammary tumors


Matthew Ellis: The Genome Institute at WashU
Genome forward oncology, how do we get there?

Retrospective somatic mutation screens within well designed sample rich clinical trials

a) Linking cancer phenotype and genotype

a) Hormone receptor ER+ve, HER2 -ve patients for breast cancer, given 3 different drugs and then surgery
ki67 in patients <10%, favorable
Find SNVs (BWA samtool snp filter) -> 136198 tomor somatic SNVs -> 30296 has somatic score>15 -> 15619 potential somatic SNVs -> tiered classification
If you knock down PIK3CA in E2 depleted cells, it gives high apoptosis
Similarly p53 mutations can be targeted using check inhibitors: these mutations can be marked for resistance
MAP3K1 somatic variations in 171 ER+ samples
The 3 commonly mutated genes explain only 50% of the patient samples

b) The druggable genome:

Mapping SNVs to candidate drugs indicates that several SNVs can be targeted. In rare cases druggable genome is possible (only that patient has the mutation)
Resistance genes can be targeted for drug therapy

You need to know the genome beforehand to know what drug to give. The other way (drugged already, then sequencing) does not work as large sample size is required.

Nidhi Bindal: Sanger UK
COSMIC database

Curated from various cancer sequencing databases (indels, substitutions, complex mutations)
Access:
a) Browse by gene
b) Browse by cancer type

Provides CIRCOS plots, tissue specific nucleotide frequency distribution, genome browser (with mutation density, genomic breakpoints, copy number annotations)

Aims to screen >1000 cell lines to investigate correlation of somatic mutations x drug sensitivity

Link to COSMIC 

Desmond Smith
Gliobastoma Multiforme

RH mapping -> triploid copy of each gene -> extra copy of gene B prevents extra copy of gene A -> fisher's exact test for all pair-wise interactions ->  example of Chr2 x chr11 interaction.

Combined RH network (3 human + dog + rat + mouse) -> single gene resolution (7.3 million edges).
About 270 times larger than HPRD and 100 times cheaper than HPRD

No. of genes x No. of edges graph : HPRD is scale-free (genes with small no. of interactions are large). Theoretically, it should be gaussian (RH network is gaussian) indicates we have saturated the possible interactions

Glioblastoma example: chr 11 (MARK2) x chr 3 (VHL)  -> RH vs glioblastoma network -> EGFR is the hub gene

Mark Boguski (Harvard)
Are clinical genomes alaready becoming semi-routine for clinical care?

PMC2009 : 6 pieces need to come together: Tech & tools, regulation, health care IT, medical education, genetic privacy and legal protections, insurance coverage & reimbursement

Mentioned Dave Dimmock's exome sequencing paper on a rare tumor on which I contributed with microarray analysis which brought up the XIAP gene

Presented a futuristic paradigm for cancer care. Also mentioned Peter Tonellato's paper who was earlier at MCW, with whom I worked.

No comments:

Post a Comment