Normalization in Metagenomics Sequencing By Total Number of Reads: Pre- or Post-Filtering and the Role of Unclassified Reads

Question

[ramin.k2013]

In metagenomics sequencing, a key normalization step involves adjusting for varying read counts across samples by dividing the abundance of taxa by the total number of sequences in each sample. Should this normalization account for the total number of reads before or after filtering out low-quality, short sequences and trimming? Additionally, how should unclassified reads be handled in this process? Should the total include only classified reads, or should unclassified reads also be considered, and why?