AI enters the clinic
Artificial intelligence is already in cars that steer themselves and in the assistants on our phones. It is also moving through medicine, more quietly and with higher stakes.
A recent study asked whether AI could surface new therapeutic targets for cancer. The work focused on molecular biomarkers that sort patients by survival, which is the first step toward treatments that match a tumor instead of a category on a chart.
Beyond the clinic
The team used PandaOmics, an AI platform, to look at gene-expression changes in rare DNA-repair disorders. They found that CEP135, a protein tied to early centriole biogenesis, is often downregulated in DNA-repair diseases that carry a high cancer risk.
They also found a sarcoma signal: high CEP135 expression lined up with shorter survival. Sorting patients by CEP135 levels let the researchers look for targets that might improve the sarcoma treatments already in use.
This is a concrete case of what large genetic datasets plus a model can do that a manual review cannot: find a protein, tie it to an outcome, and open a path for work that used to take years. The cycle from signal to a candidate therapy compresses. The more useful promise is not speed for its own sake. It is treatment that follows the genetics of one patient rather than an average.
The same class of models is not confined to hospitals. At Inteligart we put them into client operations: automating the repetitive work, tightening how a process runs, and building product experiences that notice a person without making a speech about it. That includes systems that flag a need before someone files a ticket, and support agents that can hold a complicated question without collapsing into a script. The aim is a clearer relationship with the customer and a calmer operation, not a demo reel.
More on this in our research notes.