Twice in the last month, a customer caught something wrong in our output before we did — a structure model quietly returning a bad fold, with everything downstream reasoning off it.
Both are fixed. They were catchable because we show you more than one model and don't hide it when they disagree. So we built more of that.
What shipped:
- Mutation triage — Stabilize now sorts your variant list by priority
- Five structure models, one viewer — Protenix and ESMFold2 join the toggle
- A bill of materials on every Bench protocol — the actual order list
Which mutations are worth the bench time?
A scored mutation list isn't a plan. Stabilize now sorts the list for you.
- Soluble targets — a Priority column ranking, "Top 12%"
- Membrane targets — a Low Confidence flag on the bottom bucket
- Both carry through to the CSV export
A triage order, not a ΔTm prediction. What to test first, or what to leave out.
Is this fold right, or just confident?
Five structure models, one viewer. Protenix and ESMFold2 join AlphaFold2, ESMFold, and Boltz-2 in the model toggle on Characterize, Complex, and Stabilize.
- Two AlphaFold3-class models, one of them MSA-free
- A new Mol* viewer — hover a residue to identify it
- AlphaFill's transplanted ligands, rendered in place
AlphaFold2 stays the default and still drives our annotations and stability scores — the toggle changes what you see, not what we score.
Better to see the models disagree now than after the construct.
What do I actually have to order?
Bench protocols now carry a bill of materials. Every protocol ends with what the bench needs — itemised, not implied by the prose.
- Cells, media, reagents, consumables, equipment
- Absolute amounts per litre of culture, column volume, or prep
- Each material mapped to the step that consumes it
No invented vendors or catalog numbers — the list stops where our knowledge does. Now you have the order list.
Also shipped
Filter and search on the Design and Stabilize tables, insertions and internal deletions in Stabilize (early access), and new AstraYIELD model scoring for cell-free and soluble yield.
All of it points the same way: from a screen full of predictions to the short list you take to the bench.
P.S. — our first prospective wet-lab validation is public: an Orbion-designed FGF21 construct against a scientist's. We won on developability, lost on yield. Both numbers are in the preprint.
