Task: Normalize a single-cell RNA-Seq dataset to account for sequencing depth and gene-length.
For UMI-count based protocols (like 10x) that don't suffer from gene-length biases, there are various methods developed to account for sequencing depth e.g.
For read-count based protocols (like Smart-Seq2),
SCnorm can for example be used to normalize for sequencing depth. However, read-count based protocols can suffer from gene-length biases that are not accounted for in
SCnorm. Furthermore, the method of
SCnorm expects raw counts, so I don't know if I can length-normalize the counts prior to using the method. Technically,
TPMs do account for both, but it is a global scale factor method, which has other drawbacks.
Question: How can I normalize my read-count based datasets to account for gene-length biases and sequencing depth without using a global scale factor method?