LANL and NVIDIA system screens nearly 65,000 molecules for cancer isotope therapy
The top five predicted chelators for actinium-225 are being made and tested; the team is also looking for a faster way to purify plutonium.
LOS ALAMOS — Los Alamos National Laboratory chemists working with NVIDIA have built an artificial intelligence system, the Chelatron, that designs chelators, molecules that grab and hold metal ions, according to the Lab.
Lab scientist Ping Yang leads the work. She said the approach to separating lanthanides and actinides “had not advanced since the 1970s. Until now.”
An NVIDIA reasoning model proposes molecule designs, Lab scientists simulate how each would behave, and the results shape the next round.
The first test targeted actinium-225, an isotope used in targeted alpha therapy for cancer. Across two research campaigns, the system examined nearly 65,000 possible molecules, and 679 were predicted to bind actinium as well as or better than a leading chelator used for comparison. Those are computer predictions, not lab results.
The top five candidates went to Laboratory Fellow Stosh Kozimor's group. “So now we are building these molecules and then we will test them,” Kozimor said.
Yang said the team is also evaluating replacements for tributyl phosphate, a solvent the Lab has used since the 1940s to purify plutonium. The release says a faster method could speed the Lab's work building plutonium pits, the cores of nuclear weapons.