Deep inside the human body, trillions of tiny messengers are constantly whispering to our cells, telling them to grow, divide, or heal. When those whispers go wrong, cancer can take hold. For decades, scientists have struggled to eavesdrop on these conversations — because there are a staggering 1.8 million spots on human proteins where these messengers might act, and researchers have fully understood fewer than 1% of them.
Now a team at Harvard Medical School has built a powerful new tool to change that. It's called KinoPlex, an AI-enabled map that reveals the three-dimensional shapes and biochemical environments of all 1.8 million of these cell signaling sites, then matches each one with the specific messengers — called kinases — that can interact with it. The work was published in the journal Nature Biotechnology.
To understand why this matters, it helps to know a little biology. Kinases are cell-communication molecules that work by attaching signaling tags to proteins, changing how those proteins behave. This process, called phosphorylation, controls cell growth, division, metabolism, DNA repair, and more. Mutations in kinases can contribute to cancer and many other diseases. There are roughly 500 different kinases, and each has its own requirements for the sites it can bind to.
The breakthrough here is the 3D perspective. Earlier research, much of it led by Harvard's Lewis Cantley at Dana-Farber Cancer Institute, focused on biochemical signatures. But KinoPlex is the first approach to also consider the physical shape of these binding areas. The team measured more than 100 properties around each site and discovered something surprising: in many cases, a site might have the right chemical "code" around it but still be physically out of reach of the matching kinase. In the end, KinoPlex revealed roughly 250,000 sites that had both a recognizable signature and a compatible structure.
"What we've demonstrated here is that it's not just a question of having that amino acid sequence available to be phosphorylated, but it also has to be structurally realizable," said first author David Vanderwall, a Harvard MD-Ph.D. student in the Gygi Lab.
The researchers built a scoring framework that translates detected phosphorylation sites into measurements of kinase activity, then tested it on leukemia cells — successfully picking out the exact kinase signals driving the cancer's growth. Testing across additional cancer cell lines, they could identify which kinase signaling pathways mattered most to each cell, hinting at which drugs might work best.
That's the goal: with around 100 existing kinase-inhibiting treatments and more than 90 FDA-approved drugs targeting kinases, doctors could match patients to the most effective therapy based on their specific cancer. "By analyzing a patient's particular cancer, their particular set of mutations, their particular phosphorylation state, we might be able to predict which drugs would provide the best therapeutic effect," said senior author Steven Gygi, a professor of cell biology at HMS.
The team is already partnering with oncologists at Brigham and Women's Hospital and Massachusetts General Hospital to test KinoPlex on clinical samples. And with help from the life sciences company Cell Signaling Technology, the tool is freely available to scientists everywhere. "Anyone can start applying it to their work," Vanderwall said. The atlas is open — and the next chapter of smarter cancer care may have just begun.