From our sequencing data,each read contains a cell barcode, so after alignment, reads can remember their cell originality, though we mix all reads from thousands of cells. But this problem is each read doesn't contain a molecular barcode and I don't know calculate DGE matrix. The program pipeline I use for analysis is from Drop-seq paper which is about high-through single cell RNA-seq and in that paper, transcripts for each gene in one single cell are counted by the unique molecular barcode. Do you guys know other methods to deal with single cell DGE matrix is molecular barcode is not available?
So to summarize, your question is: "Can you generate a DGE from RNA-Seq reads that have no UMI (aka XM or molecular barcode)."
Have you figured an answer yourself for this yet?
I'm certainly not an expert on the subject, but have been working through reproducing some of the results of the Drop-seq paper.
If you do have the DGE produced as a result of running through the drop-seq core computational protocol developed by James Nemesh of the McCarroll lab you should be good to go, as the DGE does not even have the XM (molecular barcodes) as part of the matrix. Each row is a gene name, and each column is the XC (cell barcode). So, if you have the cell barcode for your RNA-Seq reads, you should be able to calculate number of transcripts per cell by summing the numbers below each XC in the DGE.
If not, you may be able to write something yourself (preferably in awk) that will simply count the number of reads tagged as an exon of each gene, for each XC (using the same input as you would for the DigitalExpression program, out_gene_exon_tagged.bam). This will likely over-estimate the number of transcripts per gene for each XC, as you will not be merging reads that otherwise would have been if they had an associated XM within an edit distance of 1.
Hope that helps you out, lmk what you come up with.
