What if the cancers, neurodegeneration, and cardiac events we currently catch too late were all detectable years earlier, from a routine drop of blood? The signals are already there, the biology produces them years before anything shows on a scan. We have just never had the instruments to read them cleanly.
Every few years someone claims to have done for proteins what PCR did for DNA. Usually they have not, and the pitch disappears. Proteins.1 is the first platform we have seen that actually has a credible version of it, and it is why we are so excited to join Prateek, Harri and Tuan on their journey.
The PCR analogy
Before PCR, finding DNA in a sample meant having enough of it to detect directly. After PCR, one strand became a billion and the problem disappeared. Proteins never got that moment, because you cannot copy a protein. Every workaround since has leaned on enzymes to crank the signal chemically, which gets you part of the way and then stalls.
Why the field has been stuck
The stall has two causes. Enzymes fire whether you want them to or not, which sets a noise floor you cannot amplify your way below. And even the best modern version (Simoa's digital ELISA, trapping single beads in isolated wells) is still one-shot per bead: a binary yes/no per molecule, no way to re-interrogate, sensitivity capped by how many targets happened to bind during capture. Amplification has been chemical, one-shot, and capture-limited for forty years.
What Proteins.1 does differently
Rather than copy the protein, Proteins.1 copies the measurement. A single captured target sits on a magnetic bead, the bead parks over a semiconductor sensor array, and the same bead-analyte complex gets read a thousand times in sequence. No enzymes anywhere in the loop. Each individual read might be noise, but noise does not repeat the same way twice and a real molecule does, so stack enough reads and it stops being ambiguous. The PCR insight was simple: if you cannot make the signal louder in one shot, make more shots. Proteins.1 pulls the same move on a molecule you cannot actually copy.
The market: specialized applications
Ultra-sensitive proteomics is a proper market on its own. Research tools, drug discovery, minimal residual disease monitoring, neurofilament light chain for early Alzheimer's, high-sensitivity cardiac troponin. That business lands around EUR 200m annual revenue once properly penetrated across developed economies.
The bigger opportunity
The bigger story is that the amplification is molecule-agnostic. Anything that binds, it amplifies. Proteins, DNA fragments, metabolites, whatever you can raise a capture chemistry against. Put the right chemistry on the bead and the same tabletop device runs genome, proteome, and metabolome at once. Omics on a single box, no mass spec, no parallel sequencing stack.
The comp here is Illumina, which built a $40bn+ business owning the readout for one modality. If Proteins.1 ends up owning the readout for all of them on one platform, the TAM is comparable and nobody else is really in the running.
Technical ceiling and achievement
The ceiling here is biology. A bead still has to find a single target molecule in a small volume of sample, and the antibody still has to hold onto it long enough to be read. No chip fixes that. The move Proteins.1 has actually made is to shift the bottleneck from detection chemistry, which is where the field has been stuck for forty years, to capture chemistry, which is the sort of problem that inches forward every year as antibodies and bead surfaces get better. That is what makes it worth backing.
Validation and IP
But the evidence so far is strong: single-molecule detection in a 2025 melanoma study, independently replicated at the University of Catania. EUR 2.99m EIC breakthrough grant. US and Finnish patents granted, and international filings pending.
Competitive moat
The platform sits at the crossover of custom semiconductor sensor design, magnetic bead engineering, and microfluidic transport. The diagnostics incumbents (Roche, Abbott, Thermo, Quanterix, Olink) are strong on chemistry and biology and weak on semiconductor sensor design for biology. Closing that gap means hiring into a talent pool that barely exists inside those organisations. It takes years, meanwhile Proteins.1 have been maturing the core IP since 2020.
Why Cloudberry
Proteins.1 is close to a canonical Cloudberry bet. The platform needs custom semiconductor sensor design, precision magnetics, and integrated microfluidics, and that stack runs deep in European research institutions and supply chains. Sourcing the expertise anywhere else would be the first hard problem, before you even got to the biology.
The team
Prateek Singh invented the core technology during his microfluidics PhD and holds 13 patent families. Harri Hallila built and sold Synoste, a regulated medical device company, so he knows the hardware-to-clinic path well. And Tuan Nguyen spent six years as CTO at AKITA by Finnadvance turning organ-on-chip research into real hardware, so he has already done once what Proteins.1 needs to do next. And Soumy Mohanty spent three years at Quanterix, the incumbent, which means he knows where the current platforms actually fall short.
Co-investment partner
We are pleased to be co-investing with Lifeline Ventures, Finland's largest early-stage firm (EUR 400m Fund VI, early in Supercell, Wolt, Oura) with great experience taking startups from Finland to global success.