Also curious about differentiation the subfiles within the categories. Computer Science background, so don't know much bio/genetics. Looking at using the data for cancer research across types using topic analysis methods.
miRSeq is micro-RNA seq. Micro RNAs are a class of non protein coding
RNAs that have regulatory functions (they typically bind to the 3`UTR
of coding mRNAs and regulate that way)
mRNA, in this situation, refers to a cDNA microarray, where
pre-designed probes on the microarray surface will bind to known
target mRNAs, which may be coding or non-coding
mRNASeq is what most will loosely refer to as 'RNA-seq'. Most protocols will capture all classes of RNA species that have poly-adenylated (poly(A)) tails. RNAs that don't have these tails include ribsomal RNAs and many enhancer RNAs, but there are always exceptions to these rules.
If I were you, I would obtain the following:
miRSeq
illuminahiseq_mirnaseq-miR_gene_expression - normalised micro-RNAseq
counts over each micro-RNA
illuminahiseq_mirnaseq-miR_isoform_expression - nomalised
micro-RNAseq counts over each splice isoform of each micro-RNA
mRNA
agilent4502a_07_3-unc_lowess_normalization_gene_level -
LOWESS-normalised cDNA expression values over each gene (data from
University of North Carolina)
mRNASeq
illuminahiseq_rnaseq-gene_expression - normalised RNAseq counts over
each gene
illuminahiseq_rnaseq-exon_expression - normalised RNAseq counts over
each exon of each gene
Once you get the data, performing analyses will be the next problem.
Thank you very much for all of this info. So, those under mRNA would be the older style gene expression data that you would gather from an Agilent DNA microarray method?
I made contact with the original researcher who performed the work off of which we are working, and he said that he used the illuminahiseq_rnaseqv2-RSEM_genes_normalized dataset on that site. This is different than the illuminahiseq_rnaseq-gene_expression which you had suggested. My understanding of the differences is that RSEM is an algorithm for counting the gene expression levels and there has been some "normalization" run. Is the set he used then based on the one you suggested?
Yes, 'mRNA' in this situation does just refer to an Agilent microarray. Microarrays are still pushing their own weight, though! The last microarray that I analysed was pretty much targeting all coding and non-coding transcripts, and provides a cheaper and reliable alternative to RNA-seq, where funding may be an issue.
RNAseqv2 is different but I believe that it's based on the same initial sequencing counts (as RNAseq) but that, yes, they process it differently in an attempt to better normalise it. You may want to take a look here: https://wiki.nci.nih.gov/display/tcga/rnaseq+version+2
Between RNAseq an RNAseqv2, I would just recommend that you check the distribution of the data and then choose the appropriate statistical tests based on this. I mentioned RNAseq ('version 1') because I and others are more familiar with it. In fact, the mere presence of the term 'RNAseqv2' serves to confuse a lot of people and I think that it would have been more appropriate to just call it by the normalisation method that it uses.
Put it this way: you will not be criticised for choosing either RNAseq or RNAseqv2. On the other hand, for the microRNAseq, you would face criticism for choosing the FFPE data because FFPE is renowned as being of poor quality.
Thank you very much for all of this info. So, those under mRNA would be the older style gene expression data that you would gather from an Agilent DNA microarray method?
I made contact with the original researcher who performed the work off of which we are working, and he said that he used the illuminahiseq_rnaseqv2-RSEM_genes_normalized dataset on that site. This is different than the illuminahiseq_rnaseq-gene_expression which you had suggested. My understanding of the differences is that RSEM is an algorithm for counting the gene expression levels and there has been some "normalization" run. Is the set he used then based on the one you suggested?
Hey,
Yes, 'mRNA' in this situation does just refer to an Agilent microarray. Microarrays are still pushing their own weight, though! The last microarray that I analysed was pretty much targeting all coding and non-coding transcripts, and provides a cheaper and reliable alternative to RNA-seq, where funding may be an issue.
RNAseqv2 is different but I believe that it's based on the same initial sequencing counts (as RNAseq) but that, yes, they process it differently in an attempt to better normalise it. You may want to take a look here: https://wiki.nci.nih.gov/display/tcga/rnaseq+version+2
Between RNAseq an RNAseqv2, I would just recommend that you check the distribution of the data and then choose the appropriate statistical tests based on this. I mentioned RNAseq ('version 1') because I and others are more familiar with it. In fact, the mere presence of the term 'RNAseqv2' serves to confuse a lot of people and I think that it would have been more appropriate to just call it by the normalisation method that it uses.
Put it this way: you will not be criticised for choosing either RNAseq or RNAseqv2. On the other hand, for the microRNAseq, you would face criticism for choosing the FFPE data because FFPE is renowned as being of poor quality.