Systems Genetics of Cancer - big data and all that
Download as PPTX, PDF4 likes1,179 views
This document discusses the hype around big data and argues that theories, hypotheses, and models are still needed to make data meaningful. It presents the view that bigger data does not change the fact that data does not speak for itself and that the scientific method is not obsolete. Examples from cancer genomics research are provided to illustrate how the author uses large datasets but still relies on modeling and established scientific principles to study genetic variations in cancer, cancer evolution, heterogeneity within and between tumors, and tumor tissue context. The future of this research is outlined as integrating pan-cancer analyses of thousands of genomes and tissues to characterize patterns across cancers and correlate genomic, spatial, and clinical features.
![[The ENCODE] data enabled us to
assign bio-chemical functions for
80% of the genome.
Function = showed up in the data](https://image.slidesharecdn.com/bigdata-warwick-150316125557-conversion-gate01/85/Systems-Genetics-of-Cancer-big-data-and-all-that-14-320.jpg)
Systems Genetics of Cancer - big data and all that
- 1. Florian Markowetz CRUK Cambridge Institute www.markowetzlab.org Systems genetics of cancer Big data and all that
- 3. “With enough data and the ability to crunch it, virtually any challenge facing humanity today can be solved.” Eric Schmidt et al, How Google Works, 2014
- 4. Prof Atul Butte (Genomic Medicine, Stanford) at TEDMED 2012 “Who needs the scientific method? Vast stores of available data (…) are simply waiting for the right questions.”
- 5. Chris Anderson, WIRED.com, 2008 http://archive.wired.com/science/discoveries/magazine/16-07/pb_theory The End of Theory: The Data Deluge Makes the Scientific Method Obsolete ‘Petabytes allow us to say: "Correlation is enough.” (…) We can throw the numbers into the biggest computing clusters the world has ever seen and let statistical algorithms find patterns where science cannot.’
- 7. Karl Popper 1902 - 1994 R I P, Inductionism!
- 8. Conclusion: The hype about Big Data is built on zombie ideas
- 9. The data never speak for themselves
- 13. www.nature.com
- 14. [The ENCODE] data enabled us to assign bio-chemical functions for 80% of the genome. Function = showed up in the data
- 15. Graur et al, Genome Biol Evol (2013) “This claim flies in the face of current estimates according to which the fraction of the genome that is evolutionarily conserved (…) is under 10%.” Function = evolutionary conserved
- 16. What is biological function? A conceptual question independent of the size of your data set
- 17. Always the same science. Always the same questions. Big Data is a technical challenge, not a conceptual one
- 18. Systems Genetics of Cancer Genetic variation • In people • In tumours • In clones Phenotypic variation • Tumour subtypes • Aggressiveness • Survival
- 19. Cancer genome Evolution Cancer tissue Context Cancer genome Function Ines Wei Edith Geoff Ke Anne Joe Leon Andy Amanda
- 20. Cancer genome Heterogeneity and evolution
- 21. Clonal evolution
- 23. Intra-patient heterogeneity in HGSOC Schwarz et al, PLoS Comp Bio 2014 Schwarz et al, PLoS Medicine, 2015
- 27. Mixture model 1. How many clones are there in the sample? Data Nr of clones Size of clone Variability inside clone Parameters
- 28. Graphical Model behind BitPhylogeny Phylogeny prior Prior on local parameters Likelihood The BitPhylogeny model
- 29. FUTURE • ICGC pan-cancer analysis: 2500 genomes => 2500 trees • Characterize the 2500 trees • Correlate trees with clinical data • Infer onco-genetic progression models across the 2500 trees
- 31. DNA RNA Protein ChIP Van’t Veer et al (2002) http://ms.lbl.gov Ross-Innes et al (2012) Tumors are complex tissues
- 34. Automated image analysis Supervised classification Spatial smoothing Cell types and location H&E
- 35. Quantitative analysis of tumour composition
- 36. Spatial features of tumour tissue
- 37. FUTURE • ICGC pan-cancer analysis: 2500 genomes • Collect tissues for as many samples as possible. • Correlate tissue architecture with clinical data • Correlate tissue architecture with evolution.
- 38. Florian Markowetz CRUK Cambridge Institute www.markowetzlab.org Big Data in Genomics Thank you !