Skip to content
Science Journaling Club

INTERACTIVE MODELFIELD NOTEARCHAEOGENETICS

The Sequencing Bench

A live model accompanying “What the Garum Vat Left Behind”

Read the full field noteLaTeX source

Study
Field note
Review
Peer-edited by the club review board
Article
14 sectionsabout 20 min
Files

Everything here runs on numbers the paper actually reported: an endogenous DNA range of 4–21%, a mitochondrial coverage range of 1.3× to 197.6×, eighteen bone samples extracted, seventeen sequenced, and a radiocarbon date calibrated to two probability ranges. The first model turns the paper's own definition of endogenous DNA into a sequencing budget. The second draws random bone fragments from inside the paper's reported preservation range so you can feel how uneven that range actually is. The third just does the arithmetic of "how many years ago is cal AD X" for you, since nobody keeps that conversion in their head.

endogenous DNA % = (reads mapping to sardine) / (total sequenced reads) × 100

Model 1. The coverage calculator

Say you want a target depth of coverage across the sardine mitochondrial genome (about 16.5 thousand bases in a typical teleost fish, the usual ballpark for this kind of genome). If your extraction only yields a given endogenous DNA percentage, most of what comes off the sequencer is not sardine at all — it is bacterial, environmental, or human DNA riding along in the same tube. Lower endogenous fractions mean you have to pay for dramatically more total sequencing to hit the same target. Move the sliders and watch the bar of "reads you'd have to buy" stretch.

sardine reads needed: –total reads to sequence: –"wasted" (non-sardine) reads: –waste share of the run: –

Mitogenome size fixed at 16,500 bp. Reads-needed uses coverage × genome length / read length, then divides by the endogenous fraction to get total reads — the same relationship the paper's own endogenous-DNA percentage describes, run in the other direction. This is the club's own calculator, not a figure from the paper.

Model 2. Draw a fragment from the vat

The paper does not publish all seventeen individual sample values, only the range they fell in: 4–21% endogenous DNA, mitochondrial depth from 1.3× to 197.6×. This model draws a random fragment from inside that reported range (log-uniform, since preservation varies multiplicatively rather than by even steps) and tells you whether it would have cleared a usability bar you set. Draw a few and watch how much the outcome swings from one bone chip to the next, even though every fragment in this model comes from the same vat.

Draw
this fragment's endogenous DNA: –this fragment's mtDNA depth: –draws so far: 0cleared threshold: 0
Press “Draw a fragment” to pull a randomized sample from the paper's reported preservation range.
Success rate so far
–
Paper's actual rate
17/18 = 94.4% sequenced; 12/17 ≈ 70.6% deep-sequenced
Best draw so far
–
Worst draw so far
–

Endogenous DNA and mitochondrial depth are drawn independently, each log-uniform within the paper's reported bounds (4–21% and 1.3×–197.6×). The paper does not report a correlation between the two, and neither does this model. Club's own randomization, not real per-sample data — the true values for all seventeen samples were never published individually.

Model 3. How old is “about seventeen hundred years”?

The paper reports two calibrated radiocarbon ranges for vat one, not a single date, because radiocarbon calibration never collapses to a point. Drag the "viewed from" year and watch both ranges convert live into "years before that year." The 2026 mark is where the club's own arithmetic in the field note lands.

95.4% range (AD 84–394) is: –68.2% range (AD 162–321) is: –

Source dates: radiocarbon age 2280 ± 30 BP, calibrated by the paper to cal AD 84–394 (95.4% probability) and cal AD 162–321 (68.2% probability). The subtraction against a chosen viewing year is the club's own arithmetic.

What to take away from pushing the sliders

The first model is the whole reason "endogenous DNA percentage" gets reported at all in ancient DNA papers: it is not a quality score for bragging rights, it is a direct multiplier on how much sequencing money and machine time a sample costs. The second model is about honesty regarding what "a vat full of bone" actually means — it is not one preservation state, it is a distribution, and the paper's 4% to 21% range is the width of that distribution, not a typo. Neither model reproduces the paper's own pipeline, which used real reads mapped with real software against a real reference genome. What is here is the arithmetic underneath it, left where you can check it.