Dear Seqanswers Forum Members,
It is not clear to me how many illumina reads are required for a differential
expression experiment in a mammalian genome. I have basically 2 things to go on:
1. The statement by Anders and Huber "It can be seen that, for counts below
approximately 100", even a small increase in count levels reduces the impact of shot noise...For weakly expressed genes, in the region where shot noise is
important, power can be increased by deeper sequencing, while for the higher
count regime, power can only be achieved by further biological replicates".
2. The statement by Mortazavi et al. " A 40 million read trasnscriptome measurement provides reliable measurement of a single transcript per cell".
What I am looking for is:
1. How many reads are necessary to get reliable differential expression fora fold change of 2 (or equivalently) 1/2 if present with n biological replicates in each sample. I would estimate this for microarrays using the power of the frequentist t-test and assuming p=0.001 to compensate for false discoveries. i would do the actual analysis with Limma and Benjamin-Hochberg but I use the simpler model for the power.
2. (related question). Is there an expression for the power of the negative binomial test that shows the number of biological replicates
necessary to detect a statistically significant difference between
condition 1 with counts=k1 and condition 2 with counts=k2 to a specified alpha with a given power?
3. Is there a good rule of thumb for the number of reads for differential
expression. Intuitively, I think that it would be larger than the 40M
reads necessary to observe each transcript at least once.
I would greatly appreciate any suggestions that you may have.
Thanks and best wishes,
Rich
------------------------------------------------------------
Richard A. Friedman, PhD
Associate Research Scientist,
Biomedical Informatics Shared Resource
Herbert Irving Comprehensive Cancer Center (HICCC)
Lecturer,
Department of Biomedical Informatics (DBMI)
Educational Coordinator,
Center for Computational Biology and Bioinformatics (C2B2)/
National Center for Multiscale Analysis of Genomic Networks (MAGNet)
Room 824
Irving Cancer Research Center
Columbia University
1130 St. Nicholas Ave
New York, NY 10032
(212)851-4765 (voice)
[email protected]
I am a Bayesian. When I see a multiple-choice question on a test and I don't
know the answer I say "eeney-meaney-miney-moe".
Rose Friedman, Age 14
It is not clear to me how many illumina reads are required for a differential
expression experiment in a mammalian genome. I have basically 2 things to go on:
1. The statement by Anders and Huber "It can be seen that, for counts below
approximately 100", even a small increase in count levels reduces the impact of shot noise...For weakly expressed genes, in the region where shot noise is
important, power can be increased by deeper sequencing, while for the higher
count regime, power can only be achieved by further biological replicates".
2. The statement by Mortazavi et al. " A 40 million read trasnscriptome measurement provides reliable measurement of a single transcript per cell".
What I am looking for is:
1. How many reads are necessary to get reliable differential expression fora fold change of 2 (or equivalently) 1/2 if present with n biological replicates in each sample. I would estimate this for microarrays using the power of the frequentist t-test and assuming p=0.001 to compensate for false discoveries. i would do the actual analysis with Limma and Benjamin-Hochberg but I use the simpler model for the power.
2. (related question). Is there an expression for the power of the negative binomial test that shows the number of biological replicates
necessary to detect a statistically significant difference between
condition 1 with counts=k1 and condition 2 with counts=k2 to a specified alpha with a given power?
3. Is there a good rule of thumb for the number of reads for differential
expression. Intuitively, I think that it would be larger than the 40M
reads necessary to observe each transcript at least once.
I would greatly appreciate any suggestions that you may have.
Thanks and best wishes,
Rich
------------------------------------------------------------
Richard A. Friedman, PhD
Associate Research Scientist,
Biomedical Informatics Shared Resource
Herbert Irving Comprehensive Cancer Center (HICCC)
Lecturer,
Department of Biomedical Informatics (DBMI)
Educational Coordinator,
Center for Computational Biology and Bioinformatics (C2B2)/
National Center for Multiscale Analysis of Genomic Networks (MAGNet)
Room 824
Irving Cancer Research Center
Columbia University
1130 St. Nicholas Ave
New York, NY 10032
(212)851-4765 (voice)
[email protected]
I am a Bayesian. When I see a multiple-choice question on a test and I don't
know the answer I say "eeney-meaney-miney-moe".
Rose Friedman, Age 14