West China Hospital Transfers AI-Powered Genomic Patents to Maidi Intelligence for RMB 40 Million

Recently, West China Hospital of Sichuan University announced that four patents would be transferred through evaluation and agreed pricing, atRMB 4 million and a 2% commission on sales revenuethe proposed transaction price, transferred toMed AI Technology (Wuxi) Co., Ltd.。

It is reported that the four patents are: “A Deep Learning-Based Method, Device, and Storage Medium for Short Sequence Alignment” (Patent Publication No.: CN118351938A), “A Deep Learning-Based Variant Identification Method, Apparatus, and Storage Medium” (Patent Publication No.: CN118197420A), “A Gene Prediction or Classification Method, Apparatus, and Device for Antibiotic Resistance Genes or Virulence Factors,” and “A Method, Apparatus, and Storage Medium for Identifying Opportunistic Pathogens in Large-Cohort Metagenomic Data” (Patent Publication No.: CN118380054A).The transaction has currently passed the public notice period.

These four patents are allArtificial Intelligence and Genetic Monitoring. Through deep learning, the rational fusion of multi-dimensional feature images can potentially enhance deep classification networks, enabling the capture of richer genomic information. This facilitates downstream tasks such as short-read alignment, variant identification, pathogen detection in large-scale cohort metagenomics, and prediction of antibiotic resistance genes or virulence factor genes, thereby empowering researchers to better analyze and combat diseases.

Med AI Technology was established in 2018 and is aAn AI-driven healthcare technology enterprise dedicated to the research of artificial intelligence algorithms for medical big data.According to public information, the company was founded by a team of academicians and possesses robust comprehensive AI research capabilities through its overseas laboratories.

Centered on artificial intelligence, Med AI focuses on three major areas: first, the application of medical data science and an AI-powered expert diagnostic and treatment system platform; second, next-generation high-performance, precise imaging diagnostic hardware and software devices and service products, supported by a remote diagnosis platform; and third, the genomics AI platform DEEPOMICS.

The patents involved in this transaction will enhance the capabilities of the DEEPOMICS platform, enabling more accurate identification of the human genome, thereby advancing the development of personalized and precision medicine.

It is also worth noting that Med AI has already achieved certain accomplishments in the field of medical imaging. Its self-developed AI prototype product for thyroid imaging, ITS100, received Class II medical device certification in 2021 and has entered commercialization.

Reportedly, the ITS100 features ultra-large-scale convolutional neural networks and deep learning technologies. It enables real-time, AI-assisted diagnostic support for ultrasound-based characterization, localization, and boundary segmentation of thyroid nodules (including presence/absence and benign/malignant differentiation) as well as lesions associated with Hashimoto's thyroiditis in complex backgrounds, and generates structured artificial intelligence diagnostic reports. In the future, it is expected to reduce unnecessary thyroid fine-needle aspiration biopsies, further establishing systemic innovations in non-invasive diagnostic and therapeutic decision support and minimally invasive surgical assistance workflows.

Currently, AI applications in the medical field are already quite extensive, and AI has also demonstrated significant potential in the field of genomics.

Genomics is the collective characterization, quantitative analysis, and comparative study of all genes within an organism. It not only pertains to our understanding of diseases and their genetic underpinnings but also serves as the foundation for comprehending drug responses. Genomic research inevitably involves processing vast amounts of intricate and complex data. The ability of AI systems to capture complex patterns within large datasets and across combinations of different data modalities makes them well-suited to handle the cumbersome data in genomics research, thereby becoming a powerful tool driving the advancement of the field.

Currently, multiple research teams both domestically and internationally are attempting to create new AI tools to advance genomics research.

First, interpreting non-coding genetic variants.Compared with ordered genetic data, non-coding genetic variants are more complex and difficult to interpret. However, they are closely related to the onset and progression of diseases, and deciphering them may enable accurate prediction of disease trends.

In 2020,A research team at the University of Oxford has proposed an AI model named DeepC., megabase-scale transfer learning can be used to accurately predict topologically associating domains (TADs). DeepC can predict how the genome folds from megabase (Mb) windows of DNA sequences, thereby enabling predictions of how variations in the raw sequence affect three-dimensional genomic structure. It is reported that during the COVID-19 pandemic, DeepC was used to predict patient symptom presentation and identified single-nucleotide non-coding variants and effector genes responsible for respiratory failure caused by the disease.

Secondly, interpreting genomic data by integrating different data modalitiesGenomics is replete with cumbersome, disordered data that require multi-dimensional interpretation to be applicable in subsequent research.

In 2022,A research consortium comprising Stanford University, Harvard University, the University of Toronto, and Yale School of Medicine jointly proposed Generalist Medical AI (GMAI)This model is capable of performing a variety of tasks using very little or no task-specific labeled data, including the processing of data from medical imaging, electronic health records (EHRs), laboratory results, genomics, charts, and medical texts. In genomics research, where data are abundant, simplifying data and providing decision-making guidance are crucial components. The team stated that the GMAI model will, in the future, provide researchers with efficient and accurate multidimensional data analysis, offer clinical decision support, and serve as a novel, multifunctional medical assistant.

Finally, Interpreting Liquid Biopsy Data. Circulating cell-free DNA is abundant in body fluids and holds significant importance in both clinical biopsies and laboratory identification, making it a key focus of research efforts.

In recent years, a large number of studies on genomics AI models have also emerged in China.The “Compass Alliance,” Composed of Multidisciplinary Research Teams from the Chinese Academy of Sciences, specializing in empowering life science research with artificial intelligence. In 2023, the alliance published two significant achievements: GeneCompass, the world’s first cross-species foundational large model for life sciences; and CellPolaris, a generative model for gene regulatory networks based on transfer learning.

It is reported that the training dataset used by GeneCompass comprises over 126 million cells from mice and humans, representing the largest high-quality training dataset of its kind at the time. Furthermore, GeneCompass pioneers a dual-encoding approach for gene identifiers and expression levels, enabling more precise analysis of gene-gene interactions under various specific conditions, such as cell types and perturbation states.

On the industry front, research into AI tools for genomics is also flourishing.However, for most companies, AI tools are merely one component of their pipeline; few enterprises focus exclusively on developing genomics AI platforms.

AstraZeneca has partnered with Illumina, integrating Jarvis and other computational predictors into a broader framework to analyze large-scale multi-omics datasets for the identification of pathogenic genetic variants. Meanwhile, Genentech has collaborated with GNS Healthcare to identify and validate potential cancer drug targets by analyzing electronic health records and next-generation sequencing data. Additionally, pharmaceutical giants such as Eli Lilly, Bayer, and Pfizer have all made strategic moves in this field.

Relevant enterprises have also emerged in China, such asSenseTime, XtalPi, Insilico Medicineetc., are collaborating with AI companies to develop relevant tools that advance drug discovery. Meanwhile, internet giants such as Alibaba and Tencent have also entered the field. Currently,Tencent AI Lab is conducting research in the field of single-cell proteomics.According to publicly available information, Tencent AI Lab has launched the Single-Cell Proteomics Database (SPDB), the multifunctional algorithm scPROTEIN, and the deconvolution method scpDeconv. In the future, it will leverage its existing foundation to integrate single-cell multi-omics with large protein models, thereby advancing clinical application research.

In addition, some companies are poised to pivot their strategic focus and capitalize on emerging opportunities. Founded in 2006,Cyagen BiosciencesOriginally centered on animal model development, the company has, in recent years, leveraged its extensive foundation in genetic research to embark on the development of an AI-powered genomics platform. Cyagen Biosciences has innovatively launched the “AI+GCT” CRO service platform, which utilizes AI to assist in the design and optimization of adeno-associated virus (AAV) capsid proteins for gene therapy. This platform provides end-to-end support, covering upstream viral vector construction and development, midstream viral packaging and production as well as the generation of gene therapy disease models, and downstream evaluation of gene therapy efficacy.

Meanwhile, Cyagen Biosciences has also partnered with the Artificial Intelligence Innovation Center of the Tsinghua University Pearl River Delta Research Institute to establish a Rare Disease Data Center, providing data infrastructure and intelligent solutions to support the diagnosis and treatment of rare diseases.