A note on why TROOPIR scores compatibility on a single deterministic axis, why the methodology is trade-secret-protected rather than published, and why single-principal LLC is the correct structure for this product at this stage.
The first time a senior pharma scientist asked me to "show the math," I had to explain that the math is not the product. The output is the product. The math is the recipe. You do not get a recipe with a meal at a restaurant. You get a plate. If the plate is good, you return.
This argument is more controversial in computational biology than it should be. The field grew up inside academia, where the publication is the artifact and the recipe is the publication. The first generation of computational drug discovery companies came out of academic labs, and many of them kept that DNA: publish the method, charge for the platform, hope the platform is hard enough to operate that the method's publication does not commoditize you.
That hope has not aged well. The methods get reimplemented. The platforms compete on convenience. The unit economics drift toward open source. The only firms in computational biology that have crossed nine-figure revenue with intact margins are the ones that did not publish the recipe.
Pharma's internal decision-making works on a stack of single numbers. Is this molecule promotable to a lead. Is this lead promotable to a candidate. Is this candidate promotable to the clinic. A computational platform that returns seven probabilities and an ensemble distribution with a footnote about model uncertainty is not contributing a decision; it is contributing a meeting.
TROOPIR returns one score on one axis between zero and one. The score is reproducible to the byte. The same inputs run on Tuesday return the same number on Friday. There is no temperature parameter, no random seed, no need to "average over runs." If the regulatory team asks for the audit trail, the audit trail is the input file and the score.
"A regulator can re-run the calculation. That is the whole product. Everything else is supporting material for that one sentence."
This is not the only valid design choice in the field. AlphaFold solved a different problem brilliantly and earned a Nobel for it. Schrödinger built a decades-long franchise on physics-based ensembles. Vina is the workhorse of academic docking. Each of those occupies a real place. TROOPIR occupies the place where a number has to survive a records request. That is the place where most current methods cannot stand.
If I published the recipe, three things would happen within twelve months. A graduate student would reimplement most of it. A competitor would reimplement the rest and ship it inside their platform. The score itself, the actual artifact pharma pays for, would still work, but the moat would be gone and the licensing market would compress.
Patents alone do not solve this. Patents disclose. Once disclosed, the underlying mathematics is in the literature whether the claim is enforceable or not. The trade secret framework is what keeps the recipe out of the literature in the first place, while the patent claim sits over the top to give enforcement teeth.
This is the same structure Coca-Cola has used for a century, the same structure that protects Google's ranking signals, the same structure that protects Renaissance Technologies' Medallion strategy. It is unfashionable in computational biology, where the cultural prior is publish-or-perish. It is also the only structure that supports a million-dollar license for a non-physical good.
A single principal LLC is a structural choice with three properties. It concentrates decision-making. It minimizes the number of people who could leak the recipe. It removes the multi-founder failure mode in which the methodology-owner and the operator diverge on whether to publish.
The cost is real: no co-founder velocity, no early team's complementary skills, no big pre-seed round. The benefit is that the recipe stays with one person who built it, who signs the NDAs, who answers every inquiry, and who is on the line if a trade secret enforcement action is ever needed.
At the licensing stage we are entering, the cost is the right cost to pay. Once the engine is in commercial use at scale and the trade secret is reinforced by years of customer relationships, the calculus changes and the company can grow conventionally. We are not there yet.
If you are evaluating TROOPIR, you are evaluating outputs. You will not see the source code. You will not see the parameters. You will not see the internal nomenclature. You will see deterministic scores on targets of your choosing, delivered within five business days of NDA execution, served at a private URL gated to your evaluator credentials.
If the scores hold up under your team's clinical and chemical reasoning, we proceed to a license-tier conversation. If they do not, the engagement closes and the NDA continues to bind both sides for three years. There is no trial period middle-state. The output either earns the price or it does not.
I am told this is a hard sell into pharma. So far, it has not been.