The problem. Assigning every read to a taxon works, but it depends on a huge reference database and it can be thrown off by conserved regions that many organisms share. If the goal is a species-level abundance table rather than a label for each read, most of that work is wasted, and the shared sequence is a liability rather than a signal.

The idea. MetaPhlAn precomputes a set of marker genes that are unique to each clade, from species up to phylum, selected so their presence points to one group and nothing else. Profiling a sample then means mapping its reads against only that compact marker catalog and counting hits. Because the markers are clade-specific, the counts translate directly into relative abundances without assembly and without the whole-genome database. The approach is fast and gives estimates at species resolution.

Why it matters. It is the marker-gene counterpart to k-mer classification, and the two are the standard pair of shotgun profilers. Choosing informative features instead of using everything is the same idea that shows up across bioinformatics, here applied to taxonomy.

Verdict. A lightweight, species-level profiler that stays widely used through its later versions. Read it for the clade-specific marker design.