In-silico experiments for living cells
Run ıt dry first.
Simulate the cell before you pipette. Pick a cell, set the conditions, and see what a run costs before anything runs. Every result says whether it is checked against real data or still theory.
01 · The subject
Saccharomyces cerevisiae, S288C.
A eukaryote, so compartments are in the model from day one: nucleus, mitochondria, vacuole, peroxisomes.
02 · The model
Every compartment is accounted for.
yeast-GEM v9.1.1: 4,105 reactions, 2,748 metabolites and 1,143 genes across 14 compartments, pinned by checksum.
03 · Perturb
Take it apart. Put it back.
Knock out one gene or every gene. TP1 ranks a screen in seconds; TP2 runs all 1,143 single deletions in about a minute.
04 · Over time
Watch it change its mind.
At TP3 the diauxic shift emerges from the kinetics: glucose runs out, the cell turns on its own ethanol, growth restarts.
05 · One by one
No two cells are the same.
TP4 follows a population cell by cell: each grows at its own pace, some divide early, some die. TP5 puts them in space, as a colony or a spheroid.
The numbers
Nine cells across five tiers: from a seconds-long glance to a dosed spheroid growing in 3D.
Yeast gene essentiality on 1,481 genes the model never saw. Checked.
Between a run and its rerun from the record alone: same model checksum, seed and engine versions.
Gene and drug screens
Rank the wet-lab work
before you book the bench.
In-silico gene knockout and drug screens. Score every candidate, then take the ones most likely to matter to the bench first.
- Gene knockouts
- One gene or every gene. TP1 ranks a genome in seconds; TP2 checks each deletion against the full metabolic network in about a minute.
- Drug screens
- Dose a 3D spheroid of HepG2, HCT116, MCF7 or K562 cells. See how far the drug gets into the core and how many cells it kills.
- One word on trust
- Every ranking says whether it is checked against real data or still theory. Drug screens are theory for now.
- GPM10.94
- TPI10.76
- SSC10.68
- FBA10.65
- YEF30.60
- HXK10.40
- TDH30.35
- FBP10.25
- PDC10.24
- ADH10.01
Pick a cell. Set the conditions. See the price and the confidence before it runs.
Every request is quoted first: estimated runtime and cost in dollars, and whether that cell and tier is checked against real data or still theory.
Nothing is deducted from your allowance until you confirm. Budgets are hard stops, and a job that reaches its cap checkpoints and returns what it finished.
How a run works↗Compute tiers
Five ways to ask.
One interface.
Spend seconds to triage five hundred knockouts, minutes to watch a culture over a day, or longer to grow a colony in 3D. Every tier answers a different question.
Which of these 500 deletions look lethal? Seconds.
Does it grow, how fast, where does the carbon go? Seconds to a minute.
What do the dynamics look like over a day? Minutes.
How much does cell-to-cell variation matter? Will a few cells survive? Minutes.
Does space change the answer? Gradients, starved cores, colonies on agar.
A result is checked when it matches a measurement the model never saw. Everything else is theory until a wet lab says otherwise. How we check ↗
Run it dry first.
Drylab is invite-only while we check it against more data. Join the waitlist: invites go out in the order people joined, with a monthly compute allowance to run real experiments. Have a code? Sign in and paste it.