The missing layer between scientific systems and business models

Biopharma has no shortage of systems. It has systems for compounds, assays, omics, trials, patents, markets, forecasts and documents. Yet the assumptions that determine capital allocation still travel between them by hand.

That handoff is where scientific meaning is lost.

The spreadsheet begins too late

A conventional valuation model starts once somebody has entered probability of success, cost, timing and revenue. A portfolio model starts once somebody has decided which programs are correlated. Those numbers may be thoughtful, but the scientific reasoning that produced them is usually outside the model.

This matters because evidence classes have different empirical relationships with success. Human genetic support, for example, has been associated with higher rates of clinical progression and approval (Nelson et al.). Molecular structure can reveal liabilities or similarity. Safety evidence can make a program unacceptable regardless of its expected value. These are not footnotes to a financial model. They are part of its causal structure.

The missing product is the computation between systems

ARiDA is built around a direct sequence:

1. Resolve the molecule, target, indication and asset revision.
2. Acquire source-bound evidence and compute scientific features.
3. Translate accepted science into explicit gates or assumptions.
4. Run valuation, uncertainty and portfolio mathematics.
5. preserve the decision, approval and reopening conditions.

The molecular path is a useful example. ARiDA can calculate governed RDKit descriptors and chemical fingerprints. RDKit implements Morgan fingerprints and other established representations for molecular comparison (RDKit documentation). ARiDA does not present the fingerprint as a business conclusion. It can use a sourced druggability score to modify benchmark PoS conservatively, or use complete portfolio-wide similarity coverage to propose a correlation matrix. Those assumptions then alter rNPV, covariance and allocation through deterministic engines.

The same bridge applies to experimental strategy. Scientific uncertainty can enter EVPI, EVPPI and EVSI. EVSI estimates the expected benefit of reducing uncertainty through a proposed study design before the data are collected (methods guide). In ARiDA, that turns “we need more data” into a sharper question: which experiment is worth its cost because it could change the decision?

Integration is a decision property

Putting data sources on one screen does not integrate them. Neither does attaching citations to a report. Integration exists when a governed change in scientific evidence can change the mathematics, the recommended action and the recorded rationale.

That is why ARiDA should not be described primarily as a research workspace or evidence platform. Those capabilities are necessary. The distinctive product is the computational spine that lets science determine what to develop, license, fund, test or stop.