AI Finds Hidden Human Proteins, Revealing Unexplored Biological Functions

Researchers have used artificial intelligence, combined with advanced experimental techniques, to identify hidden and previously unexplored proteins in the human body, according to new findings published this week. The work extends the known catalogue of human proteins and provides new data on what those proteins do inside living systems.

What Happened

Scientists reported that an AI-assisted research approach successfully located proteins that had not been previously characterised in the human proteome. The research, covered by Phys.org on September 12, 2026, describes a method that pairs machine learning tools with experimental laboratory science to detect protein candidates that conventional approaches had not surfaced. Once identified, the same combined approach was used to determine the functional roles of those proteins within biological processes.

The findings were described as demonstrating that AI can contribute to protein discovery in ways that go beyond structure prediction, extending into the identification of molecules whose existence was not previously confirmed.

Background

The human proteome, the complete set of proteins expressed by the human genome, has been the subject of large-scale mapping efforts for decades. Despite significant advances, a portion of proteins encoded by the genome remained poorly characterised or undetected in standard experimental conditions. These gaps have represented a persistent challenge for researchers working in drug development, disease biology, and fundamental cell science.

AI tools have previously been applied to protein science most prominently through structure prediction. DeepMind's AlphaFold system, released in successive versions from 2020 onward, demonstrated that machine learning could predict three-dimensional protein structures with high accuracy, a task that had previously required years of experimental work per protein. The new research described this week applies a related but distinct problem: finding proteins that had not yet been catalogued, rather than modelling the shape of known ones.

What the Research Involved

According to the Phys.org report, the work relied on a combination of computational screening and laboratory validation. AI tools were used to analyse biological data and surface candidates for proteins that existing methods had not detected. Experimental science was then applied to confirm the presence of those proteins and to characterise their functions.

The report did not specify the total number of newly identified proteins or the precise experimental platforms used. Full methodological details are expected to be available in the associated peer-reviewed publication.

Why This Area of Research Matters

Proteins carry out the majority of functional tasks in living cells, including signalling, structural support, catalysis, and immune response. Proteins that have not been identified or characterised represent potential gaps in the understanding of disease mechanisms and biological processes. Researchers and pharmaceutical developers have cited incomplete proteome coverage as a limiting factor in target identification for drug discovery.

The ability to systematically locate previously hidden proteins using AI tools could reduce the time and cost associated with proteome mapping and open new lines of investigation into conditions where the underlying biology is not yet well understood.

Mathematician and science communicator Steven Strogatz, speaking separately to WIRED this week in the context of AI's broader scientific advances, described the pace of AI-driven discovery as producing results that researchers had not anticipated on this timeline, noting that the field is moving beyond what human experts can fully track or verify independently.

What Happens Next

The research team is expected to present full findings through peer review, and the methods described may be evaluated by other proteomics and computational biology groups for replication and extension to non-human biological systems.