TMM Normalization within MEDIPS ups some zero counts, is this normal?
1
0
Entering edit mode
@jenniferlneary-11265
Last seen 8.2 years ago

Hello,

I am comparing two types of methylation pull-down experiments on the same 8 genetically identical rats. (4 treatment, 4 control). I employed MEDIPS for this comparison and I chose EdgeR and TMM normalization. All seems well but we noticed in the data that sometimes (rarely) normalization raises 0 counts to a value in the RMS and RPKM columns. Is this a result of the normalization method? Is it expected? Why does it happen? Is it related to having adjacent DMRs that haven't been merged? I feel that the data is sound, I'd just like to be able to explain this. Here is an example:

DMR Start position DMR Stop position S1 raw count S2 raw count S3 raw count S4 raw count M1 raw count M2 raw count M3 raw count M4 raw count S1 RMS normalized S2 RMS normalized S3 RMS normalized S4 RMS normalized M1 RMS normalized M2 RMS normalized M3 RMS normalized M4 RMS normalized
1105201 1105300 0 0 0 0 6 6 8 8 0.22 0 0 0 0.47 0.43 0.45 0.47
1105301 1105400 0 0 0 0 7 2 8 8 0.26 0 0 0 0.53 0.35 0.51 0.51

Thank you,

Jennifer

MEDIPS • 1.2k views
ADD COMMENT
0
Entering edit mode
Lukas Chavez ▴ 570
@lukas-chavez-5781
Last seen 6.7 years ago
USA/La Jolla/UCSD
Dear Jennifer, calculation of rpkm and rms values per sample is totally independent of differential enrichment analysis between conditions. For the latter, MEDIPS transfers a count table to edgeR which is being normalised by TMM and subjected to a standard edgeR workflow. Subsequently, the result table will contain fold change, p-value, and average enrichment as returned by edgeR. Calculating rpkm and rms are different strategies and I cannot immediately see why there are non-zero rms values for windows with zero read counts. It might be due to a series of CpG density normalisation and subsequent data range shifting. In fact, MEDIPS has not much improved its concept of calculating methylation estimates since its release 2010. Instead, further development has concentrated on differential enrichment analysis between conditions (what is independent of any CpG density bias). For example, it has been shown that MEDIPS systematically underestimates methylation levels and improved methods have been proposed. If you are really interested in accurate methylation values per window and sample in addition to differential enrichment analysis, I strongly recommend to get familiar with our new package qsea available in the development branch. Please see https://www.bioconductor.org/developers/how-to/useDevel/ on instructions how to use packages in the development branch. We will be happy to assist in analysing your data using qsea and will be grateful for any feedback. Sincerely, Lukas On 09 Aug 2016, at 17:28, jenniferlneary [bioc] <noreply@bioconductor.org<mailto:noreply@bioconductor.org>> wrote: Activity on a post you are following on support.bioconductor.org<https: support.bioconductor.org=""/> User jenniferlneary<https: support.bioconductor.org="" u="" 11265=""/> wrote Question: TMM Normalization within MEDIPS ups some zero counts, is this normal?<https: support.bioconductor.org="" p="" 85966=""/>: Hello, I am comparing two types of methylation pull-down experiments on the same 8 genetically identical rats. (4 treatment, 4 control). I employed MEDIPS for this comparison and I chose EdgeR and TMM normalization. All seems well but we noticed in the data that sometimes (rarely) normalization raises 0 counts to a value in the RMS and RPKM columns. Is this a result of the normalization method? Is it expected? Why does it happen? Is it related to having adjacent DMRs that haven't been merged? I feel that the data is sound, I'd just like to be able to explain this. Here is an example: DMR Start position DMR Stop position S1 raw count S2 raw count S3 raw count S4 raw count M1 raw count M2 raw count M3 raw count M4 raw count S1 RMS normalized S2 RMS normalized S3 RMS normalized S4 RMS normalized M1 RMS normalized M2 RMS normalized M3 RMS normalized M4 RMS normalized 1105201 1105300 0 0 0 0 6 6 8 8 0.22 0 0 0 0.47 0.43 0.45 0.47 1105301 1105400 0 0 0 0 7 2 8 8 0.26 0 0 0 0.53 0.35 0.51 0.51 Thank you, Jennifer ________________________________ Post tags: MEDIPS You may reply via email or visit TMM Normalization within MEDIPS ups some zero counts, is this normal?
ADD COMMENT

Login before adding your answer.

Traffic: 374 users visited in the last hour
Help About
FAQ
Access RSS
API
Stats

Use of this site constitutes acceptance of our User Agreement and Privacy Policy.

Powered by the version 2.3.6