Count matrices correction before integrating single-cell RNA-seq datasets

Question

[berry]

Hi,

I have 3 single-cell RNA-seq datasets from the same platform (10X), same type of sample, same condition, but from different labs to integrate. When I check the genes.tsv or features.tsv files, even though the high majority of the IDs match, I see some differences. For example here "ENSG00000243485" corresponds to a different gene symbol in each dataset:

data1[data1$ENSEMBL == "ENSG00000243485", ]
>ENSG00000243485 MIR1302-2HG   
data2[data2$ENSEMBL == "ENSG00000243485", ]
>ENSG00000243485 RP11-34P13.3 
data3[data3$ENSEMBL == "ENSG00000243485", ]
>ENSG00000243485 MIR1302-10

Or here "AL627309.1" gene corresponds to a different ENSEMBL id:

data1[data1$GeneName == "AL627309.1", ]
>ENSG00000238009 AL627309.1 
data2[data2$GeneName == "AL627309.1", ]
>0 rows
data3[data3$GeneName == "AL627309.1", ]
>ENSG00000237683 AL627309.1

How would you process these matrices?

Many thanks!

[Friederike]

Can you find out which GTF file versions were used for the different samples? Presumably, they differ, and ideally, you should reprocess all samples with the same annotation file.

[berry]

Hi Friederike, thank you for your reply. I only have access to CellRanger outputs unfortunately.

[rpolicastro]

Is this from a paper or a collaborator?

[berry]

From different papers. They all used GRCh38 but I don't know about the GTF files.

[rpolicastro]

Their fastq files are likely uploaded to SRA or ENA. If they are, I would recommend rerunning them through cell ranger with the same annotation.