June 2014

Georges: 4th method multivariate regression of p<10^-6. Example of 'Results 1 NOD2' single variant test across 4 methods #ESHG

3:14am June 1st 2014 via Hootsuite

Georges: Other models - Metropolis-Hastings method ("MCMC"), a chaining of possibilities to fit data; also Bayesian method #ESHG

3:12am June 1st 2014 via Hootsuite

Georges: A statistical test of each marker. (NB: Dr. Georges has a great, precise and clear teaching style!) #ESHG

3:11am June 1st 2014 via Hootsuite

Georges: Many risk loci are in high LD; threshold r^2 0.9; reconstruction of haplotype history based upon mutations #ESHG

3:10am June 1st 2014 via Hootsuite

Georges: Use of Immunochip dataset (18K Crohns, many 10's of K UC and other); Broad uses a forward selection (biggest effect) #ESHG

3:08am June 1st 2014 via Hootsuite

Georges: 'Top' SNP ('sentinel') may not be causative; can have secondary effects etc. Try a systematic, 'conditional analysis' #ESHG

3:07am June 1st 2014 via Hootsuite

Georges: In IBD, >160 WGAS-identified Crohn's risk loci. Jostins 2012 Nature http://t.co/7KKefP7RLg #ESHG14

3:05am June 1st 2014 via Hootsuite

Next: Michel Georges (University of Liège) “Control of gene expression in disease” #ESHG

3:02am June 1st 2014 via Hootsuite

Deplancke: Credits to Manolis Dermitzakis (Univ Geneva), Sebastian Waszak in his lab. #ESHG14

2:56am June 1st 2014 via Hootsuite

Deplancke: Refers to Bing Ren's work (UCSD) about chromatin topology suggested 2013 Nature pub http://t.co/xUZPOAwl8M #ESHG14

2:55am June 1st 2014 via Hootsuite

Deplancke: VMM's show strong co-variability; do they reflect chromatin topology? #ESHG14

2:53am June 1st 2014 via Hootsuite

Deplancke: Example of UGTB17 gene locus - where one binding allele is deleted, explaining low VMM activity #ESHG14

2:52am June 1st 2014 via Hootsuite

Deplancke: PU.1 binding event variability - believe that it is the VMM rather than the minority that have variation in the PU.1 site #ESHG14

2:50am June 1st 2014 via Hootsuite

Deplancke Detail around the PU.1 motif and allele-specific binding patterns when disrupted - but how when site not variable? #ESHG14

2:49am June 1st 2014 via Hootsuite

Deplancke Referred to recent pub (Science Nov 2013: http://t.co/1V91O7zcMH ) Classify heterozygote SNPs with PU.1 motifs #ESHG14

2:46am June 1st 2014 via Hootsuite

Deplancke: 50% of QTLs are >10kb from mark; 33% of PU.1 QTLs are local; QTLs w/in the mark, the greater the effect #ESHG14

2:44am June 1st 2014 via Hootsuite

Deplancke: Observed remarkable variation of VMMs - looked at QTL's within 250kb of the variable peak; relatively low 2% QTLs #ESHG14

2:43am June 1st 2014 via Hootsuite

Deplancke: Related to gene exp variation: RNApol II correlates with the gene; but majority of VMMs not ass'd with gene exp. #ESHG14

2:41am June 1st 2014 via Hootsuite

Deplancke: Chr21 APP genes - 275kb, 23 peaks, 5 molecular phenotypes; individual cell lines observe coord. global chromatin activity #ESHG14

2:39am June 1st 2014 via Hootsuite

Deplancke: Strong enrichment of var enhancers; >90% VMMs have at least 1 H3K27Ac peak - importance of that mark #ESHG14

2:38am June 1st 2014 via Hootsuite

Deplancke: 5 sites (RBP2, H3K4me3 etc) across Chr21 example - from 78k phen-phen ass'ns, 14.5K VMMs #ESHG14

2:37am June 1st 2014 via Hootsuite

Deplancke: This graph is binary relationships - but what about higher-order complexity? Variable Molecular Models (VMMs) #ESHG14

2:36am June 1st 2014 via Hootsuite

Deplancke: Distances up to 1MB apart - H3K4me3-RPB2 100bp away - but many over 'long distances' 100kb-500kb. #ESHG14

2:35am June 1st 2014 via Hootsuite

Deplancke: Across 47 lymphoblastoid lines: looking at H3K27ac marks against H3K4m1, PU.1 etc. 2.7M pairwise comparisons #ESHG14

2:34am June 1st 2014 via Hootsuite

First up: Bart Deplankce Univ Luasanne Variation and genetic control of chromatin in humans #ESHG14

2:32am June 1st 2014 via Hootsuite