Let Drug-Resistant Patients’ Tissues Set the Problem First: Sightera Raises €3 Million to Reorder AI Drug Discovery

AI can rapidly propose large numbers of new molecules, but a structure that looks ideal to a computer may not remain a good drug once it enters the complexity of the human body. Belgian biotech company Sightera Biosciences is attempting to address this gap in reverse: first reading drug responses from the tissues of patients whose treatments have failed, then having generative AI design candidate molecules based on those biological signals.

On July 17, Sightera announced the completion of a €3 million pre-seed financing round led by Entourage, Anacura, and Qbic. The funds will be used to expand its AI platform and preclinical pipeline, recruit research and data science talent, and advance partnerships with pharmaceutical companies. The company is a spinout from the University of Antwerp and Antwerp University Hospital and is currently focused on small-molecule drugs for oncology and fibrosis.

The starting point for this approach is biological material from patients with advanced, treatment-resistant, or terminal diseases. Through hospital collaborations, the team obtains samples and builds models that include patient-derived organoids, then uses high-throughput screening to observe how different interventions alter tissue states. The resulting drug-response data then serve as the basis for the AI to design new molecules.

This differs from the common process of selecting a target first, generating chemical structures, and then validating them layer by layer. Sightera aims to begin with the biological networks already altered in drug-resistant tissues, exposing candidate drugs from an early stage to tests that more closely resemble patients’ disease states. However, although organoids can retain some tissue characteristics, they still cannot fully reproduce the human body’s immune system, metabolism, blood flow, or interactions between organs. Replacing the data source with patient samples does not guarantee that effectiveness in a model will translate into effectiveness in humans.

The financing will also advance the lead program, SIGHT001. Publicly available information states only that it is a “molecular glue” program in oncology that is progressing toward preclinical candidate selection. Molecular glues generally use small molecules to facilitate interactions between proteins that would not otherwise readily bind, potentially altering or eliminating disease-causing proteins. However, Sightera has not disclosed SIGHT001’s target, indication, animal-study results, or safety data.

At this stage, the biggest unknown remains how much quantifiable advantage the platform actually provides. The company has not disclosed the number of samples, disease types, or data quality-control methods, nor has it provided head-to-head comparisons with models trained on public data or conventional drug-discovery processes. Patient-derived data also involve batch variation, representativeness, consent procedures, and privacy governance. If the models or data are continually updated, a traceable validation framework will also need to be established in the future.

This funding therefore resembles an early bet on a research and development hypothesis more than proof that AI drug discovery works. The company describes this round as the first step toward a larger financing and says its long-term goal is to bring SIGHT001 into human trials. Before then, candidate selection, reproducible preclinical efficacy, toxicology, and manufacturing conditions remain hurdles that must each be overcome.