Parameters for 10x 5' alignment about star HOT 1 OPEN

On digging into this issue a bit deeper: some samples in the 5' dataset have 26 bp on the first mate while other samples have 150 bp.

For samples that have 26 bp, the first mate contains just the cell barcode and UMI. I had initially made the assumption that STAR would just map these as single-end reads even when using the above 5' parameters. However, when looking at the SAM flags I realized that all reads in these samples were not properly paired. I mapped these as single-end reads, and the counts improved.

For the samples that have 150 bp, I verified that the first 39 bases contains the cell barcode, UMI, and the 13 bp TSO. When these were aligned using the 5' commands above, STAR had marked all reads as properly paired but the counts were low as indicated above. However, if these samples were aligned as single-end reads, the counts improved and were close to the Cellranger.

The figure below compares the counts when aligned as single-end reads:

Please could you let me know any insight on why the samples samples with 150 bp on the first mare are getting low counts when aligned as paired-end reads?

from star.