Hit rate — the fraction of synthesized variants that pass a functional threshold — has become the default shorthand for measuring how well a protein design approach works. It shows up in papers, vendor comparisons, and internal campaign reviews. We use it too, because it is fast to compute and easy to communicate. But we have spent enough time looking at how protein engineering campaigns actually succeed or fail to believe that hit rate is a systematically misleading proxy for the thing you actually care about: the total cost of reaching a validated variant that meets your application specifications.
This post argues for a different set of campaign metrics, explains why they give a more honest picture, and walks through a specific example where two approaches with identical hit rates produced dramatically different actual costs.
The hit rate problem, precisely stated
Suppose two protein design approaches both yield a 30% hit rate for a given enzyme engineering goal — meaning that 30% of synthesized variants show activity above a threshold. At first glance they are equivalent. But consider what is left unspecified by that number:
- What was the quality distribution of the 30%? Did they cluster near the threshold, or did some variants substantially exceed it? A design approach that consistently produces high-quality hits is worth more than one that produces marginal-threshold hits at the same hit rate.
- How many sequences did you have to synthesize to reach the successful variants? A 30% hit rate out of 10 synthesized sequences (3 hits) is very different from 30% out of 200 synthesized sequences (60 hits needed) if your campaign goal is one confirmed working variant.
- How many synthesis rounds did the campaign require? A single synthesis round yielding 30% hits is cheaper than two rounds — even if the second round had a higher hit rate — because the calendar time and overhead costs compound.
- Did any of the hits actually meet all specifications, or did they meet one criterion while failing another? A "hit" that is active but insoluble, or active at low temperatures but unstable at process conditions, is a partial result that triggers another round.
Hit rate, as typically measured, does not answer any of these questions. It captures one dimension of campaign performance and ignores the rest.
The metrics that actually track cost
The three campaign metrics we think tell a cleaner story are: synthesis rounds to validated variant, total synthesis spend per confirmed hit meeting full specifications, and calendar days from design brief to validated variant.
Synthesis rounds to validated variant captures whether the design approach produces useful variants on the first round, or requires multiple design-build-test iterations to converge on something acceptable. A single synthesis round, even with a modest hit rate, is almost always cheaper than two rounds with a higher individual hit rate — because each round carries fixed costs in reagents, equipment time, and researcher attention that are independent of the number of sequences screened.
Total synthesis spend per confirmed hit is the clearest economic measure. Gene synthesis pricing has fallen substantially over the past decade, but at reasonable volumes and with current pricing, a campaign that required 200 sequences across two rounds to find a variant meeting full specifications cost 4-10x more than one that found the same outcome in 20-40 sequences in a single round. Synthesis spend scales directly with total sequences ordered, regardless of hit rate within each round.
Calendar days to validated variant is important for timeline-sensitive projects and is often the metric that project managers care about most. Calendar time includes the synthesis lead time for each round (typically 10-20 business days for gene synthesis), the expression and characterization time, and the decision time between rounds. A campaign that requires three synthesis rounds takes 30-60 business days just in synthesis lead time — before any expression or characterization is counted. The compounding of lead times means calendar time is acutely sensitive to number of rounds, not just to individual round hit rate.
A concrete comparison: same hit rate, different actual cost
Take a campaign to engineer a ketoreductase for synthesis of a pharmaceutical chiral intermediate. Two teams with access to different design approaches both ran comparable campaigns and reported 25% hit rates.
Team A used a site-directed mutagenesis approach focused on the active site residues known from structural literature to contact substrate. They synthesized 40 variants in round one and tested all of them. 10 variants (25%) showed the target stereospecificity on the substrate. Of those 10, 2 met the full specification including kcat retention and soluble expression. Cost: 40 gene synthesis units, two weeks characterization, one round. Two fully validated variants in hand.
Team B used a structure-guided combinatorial approach generating mutations at six positions simultaneously. They ordered 200 variants, received 25% activity above threshold (50 variants), but most of the 50 showed marginal activity that did not meet the kcat specification. They ordered a second round of 100 variants focused on the best-performing sub-regions of sequence space from round one. Round two yielded another ~25% hit rate (25 variants), with 3 meeting full specifications. Cost: 300 gene synthesis units across two rounds, four weeks characterization across two rounds, five weeks total calendar time. Three validated variants.
Both campaigns report 25% hit rate. The actual synthesis spend and calendar time are more than 3x different. Hit rate tells you nothing about this difference. Synthesis rounds and total synthesis spend per confirmed hit tell you everything.
What this means for how we talk about Fermvyne
We could present hit rates and stop there. But we think the more honest way to describe what Fermvyne's design approach changes is in terms of the metrics that track campaign cost directly.
The value of pre-synthesizing computational predictions is not primarily that they raise hit rate within a given round — though they do that. It is that they reduce the total number of sequences you need to synthesize to find variants meeting full specifications, and they reduce the probability of needing a second or third synthesis round. A design system that generates better-ranked candidates does not just give you a higher fraction of active variants per round; it concentrates the active variants at the top of the rank order so that a small synthesis panel is sufficient to find them.
We track synthesis rounds and total synthesis budget per confirmed hit across campaigns where we have sufficient visibility into lab results. When our predictions are working well, teams are finding validated variants in single synthesis rounds of 15-30 sequences rather than multi-round campaigns with 100+ sequences. When they are not working well — in enzyme families where our confidence intervals are wide and we say so upfront — the synthesis panel needs to be larger, and we say that too.
Making these metrics actionable
The practical implication is straightforward: before starting a protein engineering campaign, define success in terms of all three campaign-level metrics, not just hit rate. How many synthesis rounds can you afford, given your timeline and budget? What is the total synthesis budget for the campaign? What is the deadline that the calendar-days metric is tracking against?
Those answers should shape how many candidates you order in round one. If you have a hard timeline and can only afford one synthesis round, you need a design approach that concentrates functional variants at the top of the rank order — because you only get one shot. If you have budget for two rounds but a tight calendar constraint, you still want to maximize the probability of success in round one, because round two adds five to six weeks you may not have.
Hit rate will not disappear as a metric — it is too easy to compute and communicate. But it answers a different question than "did this campaign succeed efficiently?" Getting clearer about which question matters for your specific situation is the first step to allocating synthesis budget and calendar time rationally.