I have analyzed the dataset of GSE38132 from gene expression omnibus. The data is from cell line breast cancer ZR-75-1 which comprises of 9 conditions with 4 replicates for each condition making it 36 samples. I used limma R package to normalize the data (quantile normalization). I noticed a great change in a group where three replicates shows similar expression where the 4th replicate of the same sample is different from all other three. I confirmed the expression from raw data by cross checking with the probe id and found 1 replicate is different from all other three. As a double validation I checked the normalized sample deposited in NCBI-GEO and I found the same. Is this possible?
Please see the heat map first four are replicates from the same sample
Yes, for sure can this be.
Given these samples are from a cell line, such differences should not occur if the experiment was well controlled. This is very likely a technical problem and in this case I would check the cell culture (growth), labelling and hybridization quality statistics (you need the raw data files to do that) and, if there is some hint that the growth, labelling or hybridization did not work well, remove this sample from the analysis.
Jochen Wilhelm thank you for your quick reply. I have cross checked the expression pattern of these replicates in raw files, normalized file (by quantile algorithm) and normalized files deposited in NCBI GEO files. I can see a sample as an outlier (not everywhere but most of the region)
It is clear that it does not show the same hypridization pattern like the other 3 samples of the same group. The question is: why? Are there plausible technical reasons (other than the fact that the sample is "outlying") that may invalidate the data?
Jochen Wilhelm this is the paper associated with this dataset Article Research Resource: Interplay between the Genomic and Transcr...
and they do not mention anything special about the replicates ...
- Badges
- Science topic
- Similar topics
- Mathematics
- Statistics
- Sampling Design
- Sample Size