Decades of Only Glycerol and DMSO — USTC Uses AI to Open a New Molecular "Pipeline" for Cryomedicine
Release time:
2026-11-17
In the field of cryobiology and cryomedicine, there is a situation that has persisted for decades without ever being shaken: the choice of cryoprotectants is painfully narrow. Red blood cell preservation relies mainly on glycerol, while nucleated cells mostly use dimethyl sulfoxide (DMSO) — the list of available agents is vanishingly short. Both types of reagents are toxic at high concentrations, the steps for adding and removing them before and after use are cumbersome, and their protective performance is not always satisfactory. Demand for long-term preservation keeps growing in cell therapy, transfusion medicine, and tissue biobanks — yet the roster of candidate protectants has seen virtually no substantive update in decades.

What has gripped this field by the throat is the research paradigm itself: the relationship between molecular structure and protective effect is hard to judge empirically, so discovering new molecules has meant relying on massive trial-and-error experiments, one candidate at a time — slow and laborious. For precisely this reason, discovering an entirely new small-molecule cryoprotectant has long stood as an unsolved challenge in the field.
AI turns trial-and-error into ranking. Recently, the research group of Professor Zhao Gang at the School of Information Science and Technology, University of Science and Technology of China (USTC), published a study in Advanced Materials offering a new way to crack this problem. The team "transferred" pre-trained models — mature tools in language processing — onto chemical molecules: letting the models read molecules' sequence representations (SMILES), learn the correspondence between structural features and existing experimental data, then rank candidate compounds by their potential to inhibit ice crystal recrystallization and other properties, prioritizing the small fraction most worth testing in the lab. From 75,513 compound records, the team carried out tiered screening and ultimately selected a dozen or so candidates for experiments — compressing the screening scale by several thousand-fold, turning experimental validation from "needle in a haystack" into "following a map." The findings were published online on November 9.
Three levels, all validated. The value of the new route ultimately rests on experiments — and this batch of candidates delivered results at all three key levels of cryopreservation:
Ice-crystal level: multiple candidates showed the ability to inhibit ice crystal growth — one of the core sources of freezing damage;
Cell level: adding the screened molecules to a low-glycerol formulation pushed the post-thaw recovery rate of red blood cells above 91%; in a conventional slow-freezing system, recovery rose from about 95% to over 97%; human umbilical cord–derived mesenchymal stem cells, paired with an optimized formulation, achieved a post-warming survival rate of 93.6%;
Tissue level: in a rat abdominal aorta cryopreservation model, blood vessels stored with the candidate molecules showed mechanical properties and tissue structure closer to fresh tissue.
From a single discovery to a repeatable pathway. The breakthrough of this work goes beyond screening out a few molecules: it connects AI's molecular screening power with experimental validation for cryopreservation, forming a reusable development pathway for new protectants. In other words, the old "try them one by one" approach — driven by luck and labor — has become a "discovery pipeline" that can be switched on again and again. The systematic discovery of novel cryoprotectants thus finally has a method to follow, with the promise of reducing cell therapy, transfusion medicine, and tissue biobanks' reliance on traditional high-concentration protectants. The paper's co-first authors are doctoral students Sicai Li and Lingyu Shi, with Professor Zhao Gang as corresponding author.
Previous article
Previous article
Latest developments
The debut and discourse of the YinFeng Life Continuation Project not only served as an authoritative demonstration of its own ethical governance framework but also contributed practical wisdom toward establishing industry norms.
Yinfeng Foundation Honored as "2025 Public Welfare Donor" by Jinan Red Cross
Looking ahead, Yinfeng Foundation will use this recognition as an opportunity to further deepen its strategic collaboration with the Jinan Red Cross, expand cooperation in areas such as life health and emergency response, and continue to empower efforts to accelerate the construction of a "new, strong, excellent, rich, beautiful, and high-quality" modern socialist strong capital city, contributing even more Yinfeng strength to this endeavor.
The team successfully developed the CryoSIM platform, an intelligent microfluidics and deep learning-integrated system. This platform deeply integrates core technologies of deep learning and microfluidics to enable high-throughput, high-precision automated analysis of oocyte membrane permeability. It provides a novel technological tool for optimizing and advancing the clinical translation of oocyte cryopreservation techniques. Additionally, it offers an innovative practical paradigm for the application of artificial intelligence in low-temperature biomedicine and reproductive medicine.
Public Welfare Partnership: A Special Letter from the Jinan Red Cross
On the afternoon of January 30, 2026, the Jinan Red Cross presented a letter of special significance to the Shandong Yinfeng Life Science Public Welfare Foundation (hereinafter referred to as the Yinfeng Foundation).
World’s First Achievement Highlights Brand Leadership
In the future, Yinfeng Life Science Research Institute will continue to uphold its mission of "Dedicated to Medical Technology, Safeguarding Human Health." It will empower brand building with more original and pioneering scientific and technological achievements, contributing wisdom and strength to Shandong's goal of building a national regional innovation hub and promoting Chinese brands on the global stage.
Global First Ovarian Tissue Dual Activation Technology Debuts at 2025 Jinan Achievements Conference
Currently, the ovarian tissue dual activation technology has been successfully applied in clinical practice at Beijing University of Chinese Medicine Shenzhen (Longgang) Hospital, having treated over 400 patients with a treatment success rate of 70%. Over the next three years, Shandong Yinfeng Life Science Research Institute plans to use Jinan as a center to gradually expand the transformation and application of this technological achievement nationwide.