Deep Origin retrospectively ranked TCIP3 first among 17 molecular glues, compounds that bring two proteins together
Deep Origin retrospectively ranked TCIP3 first among 17 molecular glues, compounds that bring two proteins together
TCIP3 connects BCL6 to p300/CBP, enzymes that help cells activate genes. In lymphoma cells, this complex activates programs that stop growth and trigger cell death. Deep Origin tested whether its computational ranking would identify the molecule that the laboratory had already selected as its lead candidate.
On 20 July, researchers from Stanford, MD Anderson, and Deep Origin published a paper in Cell on TCIP3. In some cases of diffuse large B-cell lymphoma, BCL6 keeps genes that inhibit cell division and initiate cell death switched off. One part of TCIP3 binds BCL6, while another binds p300/CBP, bringing these enzymes to regions of DNA regulated by BCL6.
TCIP3 brings p300/CBP to genes that BCL6 keeps switched off. The enzymes acetylate BCL6 and the proteins around which DNA is wrapped at these sites. This weakens the ability of BCL6 to keep the genes inactive, allowing the cell to activate programs that stop growth and initiate apoptosis, a regulated form of cell death.
The authors’ open preprint from 2025 demonstrated this mechanism in cell lines. In a 72-hour assay using lymphoma cells, TCIP3 inhibited growth with an IC50 of about 0,8 nanomoles per liter. In other words, half of the maximum inhibition was reached at a concentration below one billionth of a mole. Control compounds containing the same chemical linker, but unable to bind either BCL6 or p300/CBP, were more than a thousand times less toxic to these cells. These results showed that the effect requires formation of the BCL6–TCIP3–p300/CBP ternary complex.
In a statement released on 20 July, Deep Origin described its computational analysis. The company constructed a ternary complex for each of the 17 compounds, simulated the motion of these assemblies, and calculated the strain in the chemical linker connecting the two parts of each molecule. This linker affects whether the proteins can maintain the required orientation relative to each other. Deep Origin ranked TCIP3 first in the retrospective analysis, but its activity had already been measured experimentally before the analysis was performed.
In the international CASP competition, teams submit predictions of protein structures that experimental researchers have already determined but have not yet disclosed. A comparable test for Deep Origin would work as follows: the model would rank compounds that had not yet been measured, the laboratory would synthesize several candidates from the top of the list, and researchers would then measure their activity.
Only Deep Origin reports that tumors almost completely disappeared in mice bearing xenografts, which are transplanted human tumors. The open preprint from 2025 describes TCIP3 as a tool compound and states that the authors had not yet evaluated its efficacy in a living mouse model of this lymphoma. The full text of the Cell version published on 20 July is not available. Even if the model successfully ranks a new series of compounds, this would demonstrate predictive performance only in the selected laboratory system. Establishing how results from cell and animal models relate to outcomes in patients would require additional evidence.