I have perhaps a naive question:
In RNA-seq with Kallisto, est_counts are generated and TPMs per transcripts can be calculated. However, to compare between sample, sleuth is used to calculate between sample normalization factors. Therefore, TPM for the individual libraries need to be normalized to be scaled to be compared.
In ATAC-seq, I have seem papers that generate coverage tracks that are normalized to RPKM (given a set bin size). This is sufficient to compare between samples. Why is there no generation of a scale factor in this case?