Linnæus

The pileup at the top is long reads landing at random positions. Depth accumulates under them; GC is only drawn where there is coverage to draw it from. The k-mer spectrum separates sequencing error (near zero coverage) from heterozygous sites (half depth) from homozygous ones (full depth) — the two peaks pull apart as coverage builds, which is how you know the run is working before any assembly exists.

The contact map is the interesting one. Before scaffolding it is drawn in the order contigs happened to arrive, which is noise. Hi-C contacts then re-order them into genome order, and the map resolves into blocks — one block per chromosome. Watch it change at the scaffolding step.

The GC × coverage plot is the first thing an assembler opens. The target genome lands in one cloud. Anything that is not the target — gut bacteria, the last thing the animal ate, us — lands somewhere else, and gets screened out and reported rather than quietly deleted.

When a species is actually funded, this panel shows that species and nothing else, and the word synthetic comes off the header.

Academic rates only. Every core facility here charges more to external academics than internal users, and more again to industry — often 1.3× to 2×. The figures above are the external-academic column where one is published.

Assembly labour is not priced. No core we found publishes a rate for assembly, curation and annotation. It is real work and real money and the line stays empty until somebody quotes it, rather than being filled with a number that would feel about right.

You buy a whole SMRT cell. A 400 Mb genome at 40× needs about 16 Gb, and a Revio cell yields 90+. Per-gigabase arithmetic gives roughly $18–27 a gigabase, but providers do not sell fractions of a cell, so a single genome in isolation pays for capacity it does not use. Pooling samples is how programmes get under that.

Programme prices are not retail prices. The Earth BioGenome figure of about $3,100 a species, and PacBio’s marketing of “sub-$300 HiFi genomes”, both describe large-scale projects with negotiated rates and shared flow cells. A one-off genome bought at list price costs more, and this page would rather say so than quote the headline.

The page promises that failures are published with the money spent. Nobody has any idea what that means until they see one, so here is the shape of it. This is an example, not a real row — nothing has been funded yet.

$310 spent, nothing gained, row marked FAILED and left on the board. The alternative — quietly re-rolling until something works — is how every science-adjacent crypto project has lied so far.