Notes 20100805 Owen Group Meeting

Jump to: navigation, search

Debate exists about best way to analyze microarrays
New RNA techniques allows us to retrieve sequences with higher fidelity
Claim made that coupling with proteomics allows us to increase fidelity even more
Transcriptome, Proteome measured relatively against some negative control
- Abundances
- Fold change kinds of measurements (result is increased two-fold etc.)
What about a complete count (absolute value)?
That would be handy for many reasons
- Systems Biology
Microarrays in this paper --- attempt to give absolute abundance quantification
Gold standard -- transcriptome value is evaluated against proteome value
- Which transcriptome platform is more accurate?
Specific technologies
- Affymetrix Exon microarrays
- Illumina/Solexa RNA-Seq
- Mass Spec: Generic -- mass spec 'everything' in the array.

General design: small chip -- "genomic focus" (a region of chromosome like a gene)
Probes on microarray designed to complement specific exons
Possible to detect alternative splicing by using different exon arrays

Fixed probe on chip designed to complement labeled target
Annealing occurs -- probes light up in proportion to the amount of transcript
Affymetrix has historically allowed the measurement of absolute abundance
Intentional mismatches used to bind suboptimally (one SNP -- catches random crummy matches vs intended target)
The fluorescent difference between mismatch cells vs intended cells is the actual abundance
- Algorithms were introduced that described mismatch probes as damaging to precision for relative abundance
- Now we use random strings with varying GC content -- this apparently has some predictive power for crummy matches
The random -GC based stuff are used with a GC-RMA algorithm that contains a number of normalization steps
The random -GC arrays were released without algorithms in reaction to the criticism to the intentional mismatch era

A number of short reads are performed and we align them with the genome in order to determine the correct sequence
RNA-seq advantages over microarrays
- Technologically agnostic-- it doesn't matter where you got your numbers, the numbers are absolute integer counts
No probing needed
- We don't need to know anything about the sequence to begin with
Question: How about sequence errors?
- If there's only one nucleotide difference for instance
- If we get many many reads on a single region, then we can estimate the background rate of sequencing error

Spectral count profile should match our expectations given known protein sequence
Enzyme that acts specifically Lysine-Arginine

relative
- gives a strange curve that is symmetrical about y = -x; aligns roughly on y = x but if absolute value taken, we get a weird sqrt(x)-like curve
- Ed: Use a two-parameter curve which does a regression fit as though it were an absolute value, but will accommodate the odd quadrants
absolute
- example shown is the expression bead chips -- we see an absolute change in concentration of x on an expression level of y -- looks sigmoidal
- where x is very low -- the proportion of the signal that is background is large: leads to larger error bars (known problem)

human brain tissue
microarray vs proteomics
- pools of brains and individual brains used
- pools required for RNA sequencing
proteomics platform and the individual microarray trials used individual brains (?)

To be fair, the correlation should be in the range [0.2, 0.5]
- this is because of the weird noise that occurs in the translation step -- going from mRNA to protein
Unclear whether the sequencing (new) or microarray (classic) method is better for abundance evaluation

Calculation of high leverage points -- would the results be the same if such points were removed?
PCR -- can be used to get abundance counts -- a known opponent
Experimental design strange: Brains? Pooled vs. Single?

Contents