This June, Tekkare's CTO Robin Sarfati joins the INFRAFRONTIER Conference 2026 in Rome, where research infrastructures, academic groups and funding bodies meet around the theme "Towards in silico modelling of health and disease." The programme puts a pointed question on the table: as computational models of disease mature, what would it take to trust them enough to reshape experimental research? Tekkare's contribution sits before that question. Not how to simulate a disease, but how to assemble the evidence a model depends on, and how to keep that evidence open to inspection.
"Augmented science is the curation and certification of data, so you can rely only on what is scientific and verified, with the human stepping in at the right moment."
Robin Sarfati · CTO, Tekkare
The problem: more information than any team can follow
Biomedical literature is growing faster than most teams can keep up with. Publications, clinical trial registrations, marketing authorisations, genomic databases and real-world pricing each follow their own calendar and their own format, and they rarely update in step with one another.
Few research groups have the resources to monitor all of these at once. In practice, teams tend to follow the sources they already know, and useful information elsewhere can go unnoticed.
Generic AI tools genuinely help with the first pass of a search, and they save real time. Their main limit is traceability: the summaries they return are not always linked back to the original record, so it can be hard to tell what was left out, or how closely a statement follows the source it came from.
When that kind of output feeds into a model, the difficulty tends to carry forward. The assumptions behind it become harder to see, and some time later it is not always easy to trace where a particular figure came from, or to confirm that it still holds.
Fragmentation adds to this. Connecting a trial record to a genomic entry, or to a health authority opinion, often takes manual alignment that not every team can afford. And when someone does that work, it tends to stay in their own notes rather than becoming a resource the next project can reuse.
The method: OIP Discovery® and the human-in-the-loop evidence engine
OIP Discovery® is the platform Tekkare built for this. Instead of writing a narrative, it returns a set of records, each one linked back to its source, so a researcher can open it, verify the source, and decide whether it belongs in the evidence base.
The search and the normalisation run automatically. The judgment stays with the researcher, who decides what to keep and why. "By curating the evidence ourselves," Robin explains, "we keep the AI from adding interpretations we never asked for. The interpretations have to stay scientific." That split lets a small team cover a wide field without handing the scientific call to a model.
The platform draws on clinical trial registries, PubMed, regulatory records from the FDA and EMA, genomic sources and pricing data across roughly fifteen countries. Records connect through shared identifiers, which is how a gene links to a trial, a trial to an authorisation, an authorisation to a price. Algorithms built with Orphanet add cross-disease links, surfacing connections between rare conditions that manual search tends to miss.
What comes out is a shared knowledge base with its sources attached, one a consortium can reopen a year later and still understand.
Application: rare disease research
Rare disease research is where this matters most, because the evidence is thin and scattered across conditions. Tekkare has been applying this approach with a rare disease research consortium, where information is especially fragmented from one disease area to the next.
From fragmented sources to shared research priorities
For a rare disease research consortium, OIP Discovery® supports project evaluation and research prioritisation across disease areas. It connects molecular data with clinical records, regulatory opinions and real-world data such as pricing, sales figures and health authority decisions, giving a shared picture of what already exists and what is worth developing.
"We don't stop at finding molecules," Robin notes. "We look at what is happening on the ground: drug prices, sales, the opinions of health authorities. That is what makes the context complete." For in silico work, that changes the starting point. A compound that looks promising in the literature may run into access barriers no paper mentions. A disease with almost nothing published may be well covered in registry or pricing records. Bringing both into view before they enter a model is what lets someone, later on, check the assumptions a simulation rests on.
Building institutional partnerships around structured evidence
INFRAFRONTIER 2026 gathers a mostly institutional crowd: research infrastructures, academic consortia, funding bodies and national research organisations. That setting fits how Tekkare works, because the approach was built for science done across institutions, where evidence has to be shared and checked by people who did not assemble it.
Robin is in Rome to present the methodology, and also to find research institutions facing the same bottleneck: turning scattered sources into a structured, traceable base that decisions can rest on, instead of starting each project from a blank page.
Tekkare's work with a rare disease research consortium shows what a technology platform and a scientific consortium can build together. Tekkare is looking for partnerships of the same kind, with institutions that want to make biomedical research traceable from the first search to the final result.
Why this matters for the field
As computational models take a larger role in drug development and research funding, the evidence beneath them stops being a detail and becomes a scientific question in its own right.
Augmented science is one answer. The platform handles the searching and the sorting, and the researcher keeps the scientific judgment. Every selection carries a record of where it came from, ready to be reopened later. Because Tekkare's infrastructure is HDS certified and ISO 27001 audited, that provenance holds all the way from the first query to the final report.
