Transcript in trans

Class Paper:
5-7 pages (with figures)
Scope is 1 aim (one main topic, can have multiple sub aims)
* Topic must utilize at least one genomic sequence in some way
Paper Outline
Specific Aim
Background & Significance
Research Description
Potential Pitfalls and Alternate Approaches
Short description of topic: due THURSDAY, March 25
Final paper due: THURSDAY, April 29
1
What contributes to the evolution of gene expression?
How many loci underlie expression variation?
Few major effectors or many minor contributors?
What are the mechanisms of expression evolution?
Relative prominence of cis vs. trans effects?
How much of expression variation has been selected for?
2
Expression QTL (eQTL) mapping
Treat expression levels as a quantitative trait.
* With
DNAmeasure
microarrays
(andreproducible
now next-gen(heritable)
sequencing)expression differences
Can
quickly
all the
can simultaneously
phenotype
(measure)
ALL
genes at once.
between two
different
parental
lines.
3
eQTL mapping in yeast spore clones
First done by Rachel Brem et al.
4
From Rockman & Kruglyak 2006
eQTL mapping in yeast spore clones
First done by Rachel Brem et al.
5
From Rockman & Kruglyak 2006
eQTL mapping in yeast spore clones
First done by Rachel Brem et al.
6
From Rockman & Kruglyak 2006
LOD threshold in standard QTL mapping (* 1 trait)
•1000 permutations
10% LOD score threshold: 3.19
5% LOD score threshold: 3.52
Challenge with eQTL mapping is that there are thousands of traits.
7
Challenge of multiple testing
Imagine doing a single t-test with p = 0.01 the significance threshold.
* at this p-value: 1 in 100 change data could be randomly generated
But if you do 10,000 t-tests and EACH has a p = 0.01 …
expect 100 positive tests to have occurred by chance
In genomics it is common to do a Multiple-Test Correction on the p-value cutoff
* Simplest is the Bonferroni correction but it is way too stringent
Divide p-value cutoff by number of tests.
eg. 0.01 / 10,000 tests = 10-6 is new cutoff
* Better methods adjust for False Discovery Rate (FDR)
(eg. Benjamini & Hochberg or Storey’s Qvalue)
Out of total set of what was called significant, how many of those
are likely to be false positives.
8
Lessons for eQTL studies
• Only ~25% of heritable expression traits can even be mapped
- on average they explain only 30% of heritable variation
• Most traits explained by many loci
- only 3% explained by 1 locus
- Alan Orr exponential QTL model: few big effectors with lots of modifiers
• Majority of traits explained by transgressive segregation
- distribution of F2 phenotypes extends
beyond parental phenotypes
- indicates many small effectors
- suggests stabilizing selection
- also consistent with epistasis
9
Lessons for eQTL studies
• Fewer traits show
directional segregation
- Phenotypic distribution of F2’s
between the parents
- Also implies many minor
effectors
- Suggests directional selection
by ‘tweaking’
10
Lessons for eQTL studies
directional 406
319
72
epistatic 583
68
438
transgressive 2093
~16% were directional
~23% showed epistasis
1640
~82% were transgressive
985
highly heritable 3546
Brem et al. 2005. PNAS
11
Local vs. Distant and cis vs. trans
Local eQTL: “near” the affected gene
Distant eQTL: “far” from the affected gene
cis effect: often taken to mean on the DNA molecule affected
trans effect: often taken to mean takes effect through the protein/RNA
Local QTL that work in cis:
ORF
TF binding site
affects transcription
3’ UTR
affects RNA stability
12
Local vs. Distant and cis vs. trans
Local eQTL: “near” the affected gene
Distant eQTL: “far” from the affected gene
cis effect: often taken to mean on the DNA molecule affected
trans effect: often taken to mean takes effect through the protein/RNA
Local QTL that work in trans:
ORF
Coding polymorphism
that affects TF activity
13
Local vs. Distant and cis vs. trans
Local eQTL: “near” the affected gene
Distant eQTL: “far” from the affected gene
cis effect: often taken to mean on the DNA molecule affected
trans effect: often taken to mean takes effect through the protein/RNA
Distant QTL that work in trans:
ORF
ORF
14
Local vs. Distant and cis vs. trans
Local eQTL: “near” the affected gene
Distant eQTL: “far” from the affected gene
cis effect: often taken to mean on the DNA molecule affected
trans effect: often taken to mean takes effect through the protein/RNA
PHYSIOLOGY
Distant QTL that work in trans:
ORF
ORF
Most trans acting effects are likely secondary responses
(distantly-acting loci are NOT enriched for TFs)
15
Local vs. Distant and cis vs. trans
Which is more prevalent?
Estimates vary:
- Brem et al papers: ~25% traits explained by local polymorphs
- other studies say close to 100%
- Many MORE individual genes explained by distant polymorphs
* but because many link to same loci, there are
fewer distantly acting loci
But … statistical challenges likely enrich for local polymorphisms:
- FDR hurdle is higher for trans acting loci
- cis (local) polymorphisms may have larger effect size
- also depends on how “local” is defined
16
Local vs. Distant and cis vs. trans
Which is more prevalent?
Using hybrid diploids and allele-specific expression
ORF-1
ORF-2
A cis acting polymorphism will affect only the allele it’s physically linked to
17
Local vs. Distant and cis vs. trans
Which is more prevalent?
Using hybrid diploids and allele-specific expression
ORF-1
ORF-2
A trans acting polymorphism will affect BOTH alleles
18
Local vs. Distant and cis vs. trans
Which is more prevalent?
Tricia Wittkopp et al. 2004. Nature.
- 29 differentially expressed genes between D. melanogaster & D. simulans:
- Measured allele-specific expression in D. mel/D. sim hybrid with pyrosequencing
28 out of 29 show cis variation in expression
16 out of 29 affected by trans and cis variation
Conclusion: cis-acting variation is more common to explain
interspecific variation
19
Local vs. Distant and cis vs. trans
Which is more prevalent?
Tricia Wittkopp et al. 2008. Nature Genetics.
- 78 genes examined (48 within, 49 between species … 16 genes overlapping)
4 D. melanogaster strains and 4 D. simulans strains
-cis regulatory effects explained more variation
between (64%) species rather than within (35%)
… argues against neutrality, since effects should occur
at same ‘rate’ over time
- compensatory cis + trans effects also more common between species
Conclusion: trans-acting variation is more common within species (over shorter
time frames) but is more likely to have more pleiotropic and deleterious effects
… trans-acting variation more likely to be removed over time
20