Introduction
Over the past decade, humanity has entered a new era where biology and computational science are deeply integrated. Generative artificial intelligence has become the core accelerating force driving the development of life sciences.
In its early stages, AI applications were primarily concentrated in data analysis and drug screening. Today, its reach has extended to the entire R&D process, encompassing target discovery, molecular design, clinical trial optimization, and precision medicine.
The successive emergence of models such as AlphaFold, ESMFold, xTrimo, and the Chai series has firmly established the concept of "AI for Science" in the field of life sciences—AI is no longer merely a tool to assist scientific research but has become the core algorithm engine for drug discovery.
Looking globally, biomedical enterprises that rely on AI as the underlying support for their research are rapidly emerging: They replace traditional test tube experiments with algorithms and rely on models to accurately predict molecular properties, completely redefining the efficiency boundaries and development dimensions of drug discovery.
Overseas biopharmaceutical companies
Isomorphic Labs leverages the algorithmic expertise of DeepMind and utilizes AlphaFold 3 to accurately predict the structures and interactions of proteins, RNA, DNA, and ligands.
The company is transforming such models into practical tools for drug discovery, making computational biology a core component of early drug screening and continuously promoting the deep integration of AI and life sciences.
Official website link：https://www.isomorphiclabs.com/
Atomic AI focuses on the core field of RNA structure research, integrating deep learning and experimental verification methods to create generative models that can accurately predict the three-dimensional structure of RNA.
This platform utilizes high-precision spatial modeling technology to achieve RNA target recognition and small molecule drug design, filling the long-standing technological gap in RNA drug discovery and making RNA a new structural entry point for AI driven drug development.
Official website link：https://www.atomicai.com/
Relation Therapeutics relies on graph neural networks and multi omics analysis techniques to create a Lab in the Loop platform, which utilizes single-cell and spatial transcriptome data to draw relationship maps of biological systems.
This model mines potential targets of metabolic, immune, fibrosis and other diseases from multidimensional data, transforms complex biological signal networks into computable research dimensions, and builds a systematic research and development path for new drug design.
Official website link：https://www.relationtx.com/
Reverie Labs builds a large-scale drug discovery model training platform with generative AI as its core, and carries out molecular structure generation and optimization work in the chemical space.
Its core technology has been acquired by Ginkgo Bioworks and will be used to create a new generation of biochemical basic models. Reverie Labs' practical exploration confirms that the algorithm system for drug development can independently become the underlying infrastructure of the industry, making models the universal engine driving chemical innovation.
Official website link：https://www.reverielabs.com/
Relay Therapeutics focuses on the core field of protein conformation dynamics in drug design, integrating molecular dynamics simulation, machine learning, structural biology, and DNA directed library screening techniques to create Dynamo ? Exclusive platform.
This platform conducts research based on the movement laws of proteins, excavates binding conformations that are difficult to identify by traditional methods, promotes the evolution of drug development from structural visualization to dynamic regulation, and comprehensively reshapes the discovery path of small molecule drugs.
Official website link：https://www.relaytx.com/
DeepLife conducts research around the core concept of "digital cells", integrating multiple omics, deep learning, and systems biology technologies to construct digital models of cell activity.
The Cell Blueprint platform is capable of generating digital twins of cells, which can accurately predict drug intervention effects and target responses; Relying on high-precision virtual simulation not only greatly improves the efficiency of target recognition and drug reuse, but also promotes the research of life systems from laboratory physical experiments to computable digital research space.
Official website link：https://www.deeplife.co/
Benchling creates a dedicated cloud platform for life science research and development, connecting the entire process of experimental notebooks, sample registration, instrument data, and project workflow, and achieving unified management of scientific research data structures.
This platform enables experimental information to be directly integrated into the AI computing system through data standardization and process automation, building a traceable and computable underlying framework for biological research and development.
Official website link：https://www.benchling.com/
AtomNet® Relying on AtomNet® Deep neural network platform conducts virtual screening in massive compound libraries to accurately predict the binding activity between molecules and target proteins.
This algorithm can quickly identify high potential candidate molecules in the early stages of drug development, significantly reducing the experimental cycle, and has become a core collaborative technology support for many large pharmaceutical companies.
Official website link：https://www.atomwise.com/
Exscientia creates the Centaur AI platform, which deeply integrates literature, patents, data analysis, and small molecule design.
This platform combines automated experimental systems to improve verification efficiency and build a full process R&D system from hypothesis generation to candidate molecule optimization; At present, multiple drugs developed through collaboration have entered the clinical stage, and the integration mode of AI and experimental automation has achieved a new balance between precision and speed in drug development.
Official website link：https://www.exscientia.com/
Anthropic launches Claude for Life Sciences exclusive model, empowering biomedical research with natural language interactive interfaces.
This model can support core scenarios such as literature review, experimental plan generation, and bioinformatics analysis. It can also seamlessly integrate with mainstream research platforms such as Benchling and PubMed, creating a fully intelligent assistance system for researchers from data processing to report output.
Official website link：https://www.anthropic.com/
Lila Sciences relies on the scientific super intelligent platform AISF ?， Establish a complete scientific research cycle system.
This platform is driven by AI to complete the entire process from hypothesis generation, experimental execution to learning feedback, combined with an autonomous experimental system to form a continuous learning architecture, realizing the automation of scientific research processes, efficient accumulation of knowledge, and dynamic optimization of strategies, creating an evolutionary computing underlying framework for scientific research.
Official website link：https://www.lila.ai/
Moderna is one of the few companies in the industry that can deeply integrate digital systems with drug production. It has built a complete digital R&D and automated production system around mRNA drug design, achieving almost full process online from sequence generation to candidate drug validation.
Our self-developed mRNA Design Studio ? The platform supports researchers to complete sequence construction and rapid feasibility verification in virtual space, combined with highly automated synthesis and detection processes, enabling the design and experimentation of candidate drugs to be completed in a very short period of time.
Official website link：https://www.modernatx.com/
Recursion Pharmaceuticals is committed to building drug development into an AI driven automated experimental system. Its Recursion OS platform integrates multi-dimensional data from cell imaging, biochemistry, and pharmacology, and relies on machine learning models to complete target recognition and molecular design.
The platform conducts large-scale experiments through automated assembly lines, and the experimental data is fed back to model training in real time, guiding subsequent experimental iterations in reverse. The entire system forms a scientific and intelligent research and development factory with self-learning capabilities.
Official website link：https://www.recursion.com/
Domestic biopharmaceutical companies
Yingsi Intelligent's Pharma.AI platform has become an industry benchmark in the field of AI drug development. The system integrates two core modules, PandaOmics target discovery and Chemistry42 small molecule design, to achieve dual module collaboration and empower the entire R&D process.
At present, the company has accumulated over 20 R&D assets that have entered the clinical or application stage, and the scale of cooperation with international pharmaceutical companies has reached hundreds of millions of dollars; The practical value of this platform lies in fully demonstrating that AI technology can truly connect the entire drug development chain from theoretical modeling to pipeline commercialization.
Official website link：https://www.insilico.com/
Jingtai Technology has turned "AI+robots" into a laboratory reality.
The subject was the first to integrate quantum physics algorithms with automated synthesis systems, building an experimental closed-loop system that can autonomously generate data, train models, and provide feedback for optimization decisions. AI models focus on molecular design and crystal morphology research, while robots are responsible for automated synthesis and detection. This innovative model is constantly breaking the boundaries between drug development and material development.
Official website link：https://www.xtalpi.com/
Wangshi Wisdom focuses on the field of small molecule drug generation, relying on geometric deep learning and molecular dynamics algorithms to construct compound structure models.
Its self-developed MolVado system can generate and optimize molecules in three-dimensional space, combined with Transformer architecture to further improve the efficiency of molecular design. The company deeply integrates algorithms into the entire process of optimizing lead compounds, allowing the compounds generated by the model to be quickly validated and implemented in experiments.
Official website link：https://stonewise.cn/
DeRui Pharmaceutical has built a full process AI drug discovery platform, integrating knowledge graph, molecular simulation, and generative modeling technologies to create a research and development system with industrial level application capabilities.
Its PharmKG ?、 Molecule Dance ?、 Molecule Pro ? The three core platforms collaborate and link the entire drug development process from target prediction to candidate drug optimization.
The oral GLP-1 receptor agonist developed based on this system has shown significant efficacy in phase IIb clinical trials, fully confirming that AI algorithms can effectively promote innovative drugs to enter the clinical research and development stage.
Official website link：https://www.mindrank.ai/
Huashen Zhiyao focuses on the core field of antibody design, creating a Helixon Design exclusive platform that integrates deep learning, structural biology, and high-throughput screening technologies to achieve precise design of multi-target antibodies and bispecific molecules.
Its self-developed model once surpassed AlphaFold2 in protein structure prediction competitions, fully demonstrating the competitiveness of China's hard core algorithms in the field of structural modeling.
Official website link：https://www.huashen.bio/
Molecular Heart is committed to automating the entire process of protein design, with its MoleculeOS platform and NewOrigin model capable of de novo generation of protein structures and automatic sampling and iterative optimization.
In this system, AI is no longer limited to auxiliary analysis, but has become an innovative builder of new molecules. The company expects this system to fully empower the fields of drug research and development, enzyme engineering, and biomanufacturing, and promote protein generation as an engineering field that can be precisely regulated.
Official website link：https://moleculemind.com/
The DrugFlow platform of Carbon Silicon Intelligence can provide full process drug design services from target discovery to lead compound optimization.
This system integrates AI technology and physical computing methods, covering core modules such as ADMET prediction, graph neural networks, and small molecule generation. It efficiently converts the computability of molecular structures into drug design efficacy, promoting drug development closer to the standardized and streamlined mode of industrial production.
Official website link：https://carbonsilicon.ai/
Yulu Qianxing will deeply integrate molecular dynamics simulation with AI technology to build the DIVAMICS platform, which can support the screening of small molecules, PROTAC, and large molecules.
This platform can accurately simulate the dynamic trajectory of protein drug binding process, conduct computational analysis in the nanosecond to millisecond time scale, and provide professional drug efficacy evaluation and optimization suggestions for candidate molecules.
Official website link：https://www.divamicsbio.com/
Zhizhiya and Yaodu focus their attention on the medical data layer.
Smart Sprout integrates multi-dimensional information such as patents, drug pipelines, and literature to provide target intelligence mining and risk analysis services, helping pharmaceutical companies accurately control their research and development direction in the early decision-making stage.
The platform built by Yaodu covers the entire process data system from target intelligence to clinical trials, and synchronously constructs a global compound, registration, and patent exclusive database, providing solid data support for drug development trend analysis.
Official website link：https://www.zhihuiya.com/ https://www.pharmacodia.com/
DeepZhiyao focuses on the core process of document processing and application in drug development, and has launched two intelligent systems, Translation-X and Writing-X, which can automatically complete professional tasks such as clinical protocol translation and application document writing.
The company will deeply apply generative AI to the automation of scientific research texts, relying on its self-developed multi-agent architecture and massive accumulation of medical text data, transforming traditional language based work into quantifiable and computable standardized links in the R&D chain, greatly improving document processing efficiency and compliance.
Official website link：https://www.dip-ai.com/
The Medical Magic Cube and Moentropy Medicine have further expanded the data dimension.
PharmaGO® of the Medical Magic Cube; The database gathers the full lifecycle information of drugs from research and development, registration to market access, and standardizes the data through natural language processing (NLP) and machine learning techniques, providing a high-quality data foundation for drug research.
On the basis of this database, MoEntropy Medicine further integrates multi-source medical data resources, connects with large models to build an intelligent knowledge exchange system, efficiently activates the value of data, and makes data truly the core fuel for intelligent drug development.
Official website link：https://www.pharmcube.com/ https://www.pharnexcloud.com/
summary
Overall, overseas AI + biopharmaceutical companies tend to focus more on technological breakthroughs, while domestic enterprises are more focused on industrial implementation, with significant differences in their development paths.
At the level of strategy, overseas companies tend to delve deeply into a single technological direction.For example, Isomorphic Labs focuses on protein structure prediction, Atomic AI specializes in the RNA field, and Relay Therapeutics focuses on protein dynamics. The advantage of this model is that it can build a high barrier technological moat, but its shortcomings are also very obvious - the cycle of technology transformation into products is long, requiring sufficient capital and time to complete the verification loop.
On the technical route, overseas accumulation in the fields of AI basic models and biological structure modeling is more systematic.The open-source protein language model ESM3 launched by the Meta Protein team, as well as the protein models Chai-1 and antibody design models Chai-2 developed by Chai Discovery supported by OpenAI, are leading the trend of scientific research and industrial development in this field worldwide.
At the same time, the United States has a mature and complete venture capital ecosystem, with tech giants, pharmaceutical leaders, and top venture capital firms willing to bet on high-risk technologies and provide large-scale financial support. This creates long-term high investment conditions for startups, enabling them to conduct extensive technological exploration and continuously expand the technological boundaries of the industry.
Most domestic AI+biopharmaceutical companies focus on the path of "platformization+industrialization",Covering the entire process from target discovery, candidate molecule design to clinical advancement, the core goal is to shorten the cycle from algorithm validation to commercial implementation. In this process, domestic enterprises fully leverage their cost advantages and R&D speed advantages, relying on the ecological dividends driven by capital and policies to accelerate the process of industrial landing.
In the field of data infrastructure, Benchling is used as the industry benchmark overseas, while in China, there are four companies, namely Zhizhiya, Yaodu, Pharmaceutical Magic Cube, and Yaorongyun, which are deeply and thoroughly engaged in their respective sub sectors. Although data services may not be as topical as algorithm technology, they are the core underlying support of the AI for Science ecosystem - without high-quality data, even the most powerful algorithms cannot be effective.
