Ukhvat

Drug-stabilizing proteins from NISE, and what they mean for senolytics

30 June 2026· vEDva93Q

Harvard and Dana-Farber's NISE loop designed proteins that kept over 99% of an anticancer drug in active form for 50 hours at physiological pH; senolytics are among the future applications where this molecular control matters.

At physiological pH 7.4, exatecan (an anticancer drug from the camptothecin class) quickly shifts into an open lactone-ring form; the opened form enters cells less effectively and is weaker as a drug. EPIC(Q51N/M97L), a protein designed from scratch by the neural-network loop NISE, kept more than 99% of the drug in its active closed form for 50 hours, shielding the unstable lactone ring from hydrolysis.

NISE (neural iterative selection-expansion; Harvard and Dana-Farber) is a closed loop of two neural networks. LASErMPNN selects an amino acid sequence for a protein scaffold that houses the drug molecule. RoseTTAFold-All Atom or Boltz-2 then predicts the three-dimensional complex. The best candidates advance to the next round. For exatecan, the team synthesized DNA for four proteins, expressed them in E. coli, and tested binding. All four bound the drug. The baseline EPIC variant gave a Kd of 0.12 µM (Kd, dissociation constant: lower means tighter binding). Two amino acid substitutions proposed computationally, with no new experiments, improved binding roughly 100-fold, to 1.2 nM. For apixaban (an anticoagulant), APEX reached a Kd of 80 pM, comparable to apixaban's natural target, factor Xa, in a protein about three times smaller by mass. Crystal structures showed the protein largely adopted the predicted shape and held exatecan in the predicted pocket. Published June 24 in Nature by Benjamin Fry, Kaia Slaw, and Nicholas Polizzi.

Senolytics (drugs that clear senescent cells), anticancer drugs, cell therapies, and delivery systems all hinge on controlling molecules after they enter the body. If a protein can be designed around a chosen molecule, tested in a small batch, and matured with a few mutations, you can build custom proteins that shield a drug from degradation or hold it firmly in place.

The designed proteins have not been tested in animals or humans; Dana-Farber state this explicitly. The current result is lab-stage validation of binding, specificity, structure, and chemical protection of exatecan from hydrolysis.