Biology of Genomes: Thur evening tweetable talks

Price – Mixed Model association

Increase power by assuming a mixture of two normal distributions instead on 1 normal – sounds like just adding free parameters not like adding more biology. Can you give some more intuition?
PCA
- fast PCA
- if just interested in the top eigenvector there are faster ways.
- can multiple by a random KxN matrix to pick out top K eigenvectors
- agree with other PCA methods such as PLICK
top 10 eignevectors for 55,000 americans, clustering in lesss than an hour. Get expected ethnic clustering on major eigenvectors
- also separate Irish / Northern European, and Eastern European in the 4th and 3rd eigenvectors
ID alcoholism resistance gene in euro populations (previously ID’d in asain populatgions)

germ line de novo mutations
- errors in DNA replication
- DNA breakdown
- higher in M than F
- higher in older males
detect germ line de novo mutations by sequencing parents and child at frequency ~50%
Damonda dataset (in cattle) 150 pairs of parents
- 150 probands 2/parents
- > 5 grand offspring per proband
detect mutations (6 to 8) incomplete mociacism, 28% of detected mutations occurred in fetus rather than parental gametes.
detect similar frequencies for female proband as with male proband.
there exists mosaicism in germ line of parents
early cell cleavage divisions are more error prone (rapid replication issue?)

large functional and molecular variability
- reported due to genetic abnormalities
- cell type of origin
- culture conditions
- etc
are there genetic effects on variability?
HipSci – goal: create cell bank of stem cells
- skin biopsy or blood samples from 845 healthy people and 111 rare diseasess
- reprogrammed into iPSCs
- validate with transcriptome profiling, compare against previous data.
- pairwise cnv
- assay differentiation ability into germ layers (by expression markers)
- passage 20: assay transcriptome, proteome, chIP-seq histones, methylation pattern,
- then cell bank.
sources of phenotypic variation
- batch, donor, gender, medium, passage number, unexplained residuals
expression level: 25% variability is by batch effects. a little by donor. medium some effect. more than 50% residual.
methylation, Nanong Oct4, sox2 etc imaging – dramatic batch effect variation
by ‘explains most of the variance’ you mean makes the 3rd most signficant contribution after residuals and batch
1000s of quantitative trait loci (eQTLs and methylation QTLs)
this common variatino affects important loci for pluripotency and development
CD14 is reported as one of t he most epigeneticcally variable genes. 40-80% of variation explained by donor
most variable fraction of lines in methylation have greatest fraction of variance explained by donors.

are any of your donor affects related to your healthy vs. rare genetic disease patients?

ID strong signs of selection on innate immunity genes.
regulatory variants explain some of these differences.
200 participants in Beligium, 100 African 100 European descent
isolate monocytes
ID 2000 genes in infulenza response ,350 in antiviral response, 546 in inflammatory response
MHC response stronger in Europeans, TLR inflamation stronger in Africans
Balanced allelic expression vs allele favored
- allele specific events for ~1/3rd of genes
Allele specific expression QTLs indicate allele specific regulatory SNPs
- found ASE for ~40% of testable genes.
- regulatory SNPs identified in 98% of the cases which had >5% of the individuals

map allele specific response QTLs?
do allele specific response QTLs overlap iwth signatures of postive selection?
- some do: e.g.: ORMDL1 in Europeans
- LEPROT in Africans.

are there no differences in the prognosis of response to infection between African and Europeon