Belgian Biotech Sightera Raises €3M to Train AI Drug Discovery on Real Patient Tissue
Most AI drug discovery companies train their models on public chemical databases or generic biological datasets. A new Belgian startup is betting that approach misses the point entirely, and it has raised fresh capital to prove a narrower, harder‑to‑replicate alternative works better.
Sightera Biosciences has closed a 3 million euro pre‑seed funding round led by Entourage, Anacura, and QBIC. The company is a spin‑off from the University of Antwerp and Antwerp University Hospital, and it is building what it describes as an AI‑native drug discovery platform focused specifically on oncology and fibrosis, two disease areas notorious for how often promising lab results fail to translate into working treatments in real patients.
The distinction Sightera is drawing centers on where its training data actually comes from. Rather than relying on public or generic datasets the way many AI drug discovery companies do, Sightera trains its models on proprietary data generated from patient‑derived biological samples, collected specifically from individuals with advanced, therapy‑resistant disease. Those samples are used to build preclinical models, including organoids, small lab‑grown structures that closely mimic how human tissue actually behaves, which in turn generate large‑scale datasets showing how real disease biology responds to different drug candidates.
That grounding in patient biology shapes how Sightera's AI platform actually designs new molecules. Instead of generating drug candidates based primarily on chemical properties and then testing whether they happen to work biologically, the company's models design small molecules directly from biological response data observed in patient‑derived systems. The bet underlying that approach is straightforward even if the science is not: drug candidates designed around how real diseased tissue actually responds are more likely to survive the transition from laboratory promise to clinical success, a transition where the vast majority of drug candidates still fail regardless of how sophisticated the AI that designed them was.
This distinction matters enormously in an industry where failure happens late and expensively. A significant share of drug candidates that perform well in early lab testing still fail once they reach human clinical trials, in large part because the models and cell lines used during early development do not fully capture the complexity of how disease behaves in an actual patient. By anchoring its AI models in data drawn from therapy‑resistant patients specifically, rather than more general laboratory conditions, Sightera is positioning its platform to potentially catch some of those failure signals earlier in the discovery process rather than after years of investment.
The new funding will go toward several priorities at once. Sightera plans to expand its AI‑powered drug discovery platform more broadly, accelerate development of its broader preclinical pipeline, and push its lead program, a molecular glue therapy targeting oncology, toward the stage of preclinical candidate selection. Molecular glues are a class of small molecules that work by forcing two proteins to bind together in ways that would not normally happen on their own, often to flag a disease‑causing protein for the cell's natural disposal system. The approach has drawn significant industry attention in recent years as a way to target proteins that have traditionally been considered undruggable using conventional small molecule methods.
Beyond advancing its own pipeline, Sightera also intends to use the capital to deepen strategic partnerships with pharmaceutical and biotechnology companies, a common path for early‑stage techbio startups looking to validate their platform technology against real‑world drug development problems before committing to running their own programs all the way through clinical trials. The company also plans to grow its research, AI, and data science teams to support that expanded scope of work.
Sightera's emergence adds to a broader pattern across European biotech, where university‑affiliated spinouts are increasingly positioning themselves around proprietary, hard‑to‑replicate biological datasets rather than competing purely on algorithmic sophistication. As more AI drug discovery companies enter the market using similar large language model and generative AI techniques, access to unique, clinically grounded data is emerging as one of the clearer ways smaller, earlier‑stage companies can differentiate themselves from both larger pharmaceutical incumbents and better‑funded AI‑first competitors working from more generic datasets.
