The reasoning system.

What this system actually does

The reasoning system is the part of our platform that takes a large, messy body of biological evidence — peptides, receptors, signaling cascades, master regulators, disease states, outcomes, and the directional, typed relationships between them — and organizes it into a structure a computational model can reason over explicitly, rather than holding it informally in one researcher's working memory.

Two design choices make this possible. First, the underlying graph is embedded in a hyperbolic space rather than a flat, Euclidean one — a geometry that naturally accommodates the exponential branching of a real biological hierarchy without running out of representational room at the periphery, the way flat layouts do. Second, the reasoning layer on top of that embedding uses graph neural network message-passing that explicitly tracks relationship polarity — the difference between a peptide that activates a pathway and one that suppresses it — so that an activating relationship and an inhibiting one between the same two biological entities are never collapsed into a single, ambiguous signal.

Why the geometry isn't decoration

We've written a dedicated, deeper explainer on why hyperbolic space specifically — including the published academic research the choice is grounded in, and a live, clickable visualization of the geometry itself. It's worth reading if you want to understand why this isn't an aesthetic choice dressed up as science.

What the reasoning system outputs, and what it doesn't

Given a candidate biological question, the reasoning system produces a ranked set of hypotheses — connections between a signaling molecule and a possible outcome, weighted by how well-supported the connecting evidence is and how central the intermediate biology is to the broader signaling network. It does not produce a diagnosis, a treatment decision, or a claim of efficacy. Every output is designed to be traceable back to the evidence that supports it, reviewed by a human before it goes anywhere near a laboratory bench, let alone a patient.

This is the layer every other system in our stack ultimately depends on. The carrier (exosome) manufacturing and therapeutic-regulation systems both take their individualized configuration from what this system maps; the bio-processing and pathogen-reduction systems exist to make sure the physical evidence feeding it hasn't already been degraded before it arrives.