genome assembly

Imagine taking ten identical copies of a very long book, running each through a paper shredder, and then mixing all the scraps together. None of the scraps is numbered, and many look alike. Your job is to reconstruct the original book in order, using only the fact that wherever two scraps share the same run of words at their torn edges, they probably belonged side by side. That puzzle, scaled up to billions of DNA letters, is genome assembly.

Sequencing machines do not read a chromosome end to end. They read enormous numbers of short, overlapping fragments called reads. Genome assembly is the computational step that stitches these reads back into the long, continuous sequence they came from. Software looks for overlaps — places where the end of one read matches the start of another — and merges them into longer stretches, the way you would line up scraps with matching edges. Because every region is read many times over (its coverage), the overlaps are dense enough to find the true order, and reading each spot repeatedly also lets the program vote out random sequencing errors.

Assembly is hard precisely because real genomes are full of repeats: long passages that appear in many places, like a sentence printed on a hundred different pages. A short read landing inside a repeat could have come from any of those copies, so the assembler cannot tell where it belongs and the reconstruction stalls or collapses into tangles. This is why modern projects mix short, accurate reads with long reads that span across repeats, and why no assembly is ever perfectly finished — most published genomes are drafts with gaps, and a truly gapless, telomere-to-telomere human genome was only completed in 2022.