Biology‑native data push
DistilINFO argued for a 'biology‑native' data infrastructure that preserves biological meaning through multimodal datasets, agentic AI workflows and closed‑loop lab automation rather than forcing biology into generic enterprise schemas. The piece says that keeping sample lineage, protocol versions and assay context intact is central to making data usable for iterative experimentation and model building. (distilinfo.com)
Drug discovery groups are arguing that biology data should be stored in a way that keeps the experiment attached to the sample, not flattened into generic business tables. (distilinfo.com) DistilINFO made that case in an April 13, 2026 article on artificial intelligence drug research, saying useful systems need to preserve sample lineage, protocol versions and assay context across multimodal datasets. The piece said that setup matters more as companies try to link wet-lab experiments, software agents and automated lab equipment in one loop. (distilinfo.com) In plain terms, lineage is the record of where a sample came from and what happened to it, like a package tracking log for cells, tissues or compounds. Protocol versions are the exact lab instructions used at each step, and assay context is the surrounding detail on how a measurement was made. (nature.com) Those details have become more important as drug research shifts from single data types to multimodal inputs, where one model may combine images, gene activity, chemical structures and lab readouts. Therapeutic Data Commons described that push in its Commons 2.0 update, which was built to unify single-cell biology, biochemistry and drug effects in shared benchmarks for multimodal models. (biorxiv.org) The argument is also showing up in product and investor pitches. Bessemer Venture Partners wrote on April 12, 2026 that “biology-native data,” agentic workflows and lab automation feedback loops will define the next generation of biotech infrastructure. (bvp.com) Drugmakers and platform companies are already building around that idea with systems that feed new experiments back into the software stack. Recursion said its Recursion Operating System sends each data point collected from target identification through clinical trial enrollment back into the platform to improve performance, and said its automated high-throughput labs help generate one of the largest relatable datasets in pharma. (recursion.com) Benchling, which sells software for research and development teams, has built products around structured scientific records rather than generic enterprise documents. Its engineering team wrote that assay results are among the most common schematized items in Benchling, underscoring how heavily life-science software now depends on preserving experimental structure. (benchling.engineering) Data plumbing vendors are pushing the same connection between instruments and scientific records. TetraScience’s Benchling pipeline documentation says the integration moves instrument and experimental results into Benchling through data pipelines, a step aimed at keeping machine output tied to the underlying experiment instead of stranded in separate files. (developers.tetrascience.com) Open-source groups are making a similar pitch from outside commercial biotech. LaminDB describes itself as a lineage-native lakehouse for biology that supports biological file formats, registries and ontologies, using database language to solve the same problem of keeping scientific meaning attached to data. (github.com) The near-term test is whether these systems can make artificial intelligence models more useful in real lab cycles, where a changed reagent, a revised protocol or a mislabeled sample can alter the result. The companies that can keep those details intact are the ones trying to turn biology data into something a model — and a robot — can use repeatedly. (distilinfo.com)
