genome annotation

Suppose someone hands you a three-billion-letter book printed with no spaces, no punctuation, and no chapter titles — just an unbroken stream of letters. Knowing all the letters is not the same as knowing what the book says. You still have to find where the words and sentences begin and end, mark which passages are meaningful and which are filler, and label what each part does. Doing that for a raw genome sequence is genome annotation.

Annotation is the process of finding and labelling the features inside a sequence: where genes start and stop, which stretches are exons and introns, where the regulatory signals like promoters sit, and what each gene's protein product is likely to do. Computers do the first pass, scanning for tell-tale patterns — start and stop codons in frame, the splice signals at exon boundaries, statistical signatures that distinguish coding from non-coding DNA — and then leaning heavily on comparison: if a stretch closely matches a gene of known function in another organism, that function is provisionally transferred. Experimental data, especially RNA-seq showing which parts are actually transcribed, anchors and corrects the predictions.

Annotation matters because a bare sequence is almost useless without it, yet it is also one of the most error-prone and never-finished parts of genomics. Gene predictions are routinely wrong at the edges, short or unusual genes get missed, and function labels transferred from other species can be mistaken or propagate the same error across many databases. This is why annotations carry version numbers and confidence levels, why a gene's stated function is a hypothesis rather than a fact, and why the human gene count was overestimated for years before settling near 20,000 protein-coding genes.