Muda Biology's Oral LDH Inhibitor MP-5342: Starving Abnormal Immune Cells

English text is being prepared.

English text is being prepared.

[Editor's Note] The Prix Galien USA, known as the "Nobel Prize of the pharmaceutical world," announced its 2026 award winners on October 29.

Notably, innovative companies incubated by XtalPi have been shortlisted for this award for two consecutive years. Last year, Signet Therapeutics was nominated for "Best Biotechnology Product" with SIGX1094, the world's first targeted drug for diffuse gastric cancer; this year, Liman Biotechnology and MetaNovas Biotech were both nominated for "Best Startup" — the former, with XtalPi's R&D support, was nominated for its metabolism-enhanced CAR-T therapy at one-thousandth of the conventional dose, while the latter collaborated with XtalPi to develop MP-5342, the world's first oral LDH inhibitor, targeting the global autoimmune market worth over one hundred billion US dollars.

In the spring of 2021, Xu Ke of Weill Cornell Medicine received two invitations. Two renowned American professors in the immunology field, each having founded an immunotherapy company, both invited him to join as a co-founder.

Generous compensation, a clear direction, everything ready — and Xu Ke declined.

"Why don't I go do my own thing?"

At the time, in the joint training program between Cornell and Memorial Sloan Kettering Cancer Center, Xu Ke had already spent nearly eight years studying the intersection of immunology and metabolism. He had a set of data on how immune cells change their metabolic basis before launching an attack. In his view, the potential of this data was greater than any offer from an existing company.

Xianyu Anjin, CTO of MetaNovas Biotech, was then still serving as Director of Data Science at PatientsLikeMe, the largest patient data company in the United States, with an extremely high salary. But it didn't take much effort for Xu Ke to persuade Xianyu Anjin: "He took one look at my data and knew what a great thing it was."

The two hit it off immediately and decided to return to China to start a business. XtalPi was one of their earliest investors; after returning to China, the two settled directly in XtalPi's office building, becoming innovation partners "upstairs and downstairs."

The immunometabolism track Xu Ke targeted is still in a stage of continuous exploration, even abroad. There are many kinds of enzymes in metabolic networks, but developing new drugs targeting immune cell metabolism still faces challenges such as selectivity and safety.

Metabolism is too fundamental; intervening in it means pulling one hair and moving the whole body, and safety risks are hard to predict. But Xu Ke has been deeply engaged in this field since 2013, and he observed a key phenomenon: some immune cells involved in autoimmune diseases undergo "metabolic reprogramming" during abnormal activation — increasing glucose utilization to support cellular functions.

"Bad cells have to eat their fill before they can do bad things," Xu Ke said by way of analogy. MetaNovas Biotech hopes to regulate abnormal immune responses by intervening in specific metabolic steps, exploring new treatment approaches for autoimmune and metabolic diseases.

MetaNovas' oral small-molecule LDHA inhibitor MP-5342 is equivalent to "starving" the bad cells of the immune system: by inhibiting the glycolysis pathway, it reduces the metabolic activity of immune cells and returns immune responses to homeostasis, rather than crushing the immune system to rock bottom like traditional immunosuppressants.

This "starvation therapy" brings a safety window of more than 600-fold, far exceeding the industry's conventional safety threshold (typically >40-fold). Normal cells do not usually depend on this pathway; only pathological cells in a stressed state are highly dependent on it. By contrast, Humira (adalimumab), which once had annual sales approaching 20 billion US dollars, as well as current JAK inhibitors, carry boxed warnings — risk of serious infection, and even cardiovascular risk.

MetaNovas did not start with psoriasis or atopic dermatitis, which have the largest patient bases, but chose inflammatory bowel disease (IBD) first. Existing drugs for this disease generally face the dilemma of a clinical remission rate below 40%; severe patients may need to go to the toilet 20 times a day, and may even develop complications such as intestinal perforation. In addition, areas with unmet clinical needs such as primary biliary cholangitis and multiple sclerosis are also on the expansion list.

Based on the same underlying biology of immunometabolism, MetaNovas Biotech has discovered a series of entirely new targets of metabolic proteases; these targets sit at key nodes of the metabolic network and have the potential to expand to a variety of different indications. In addition to autoimmune diseases, MetaNovas is also advancing META-001-PH, a First-in-Class oral small-molecule inhibitor for primary hyperoxaluria — a rare inherited metabolic disease that often occurs in childhood, in which oxalate continuously accumulates in patients' bodies, and most patients progress to kidney failure before adolescence. Xu Ke told us that there is currently no specific drug for this disease in China.

In its R&D path, MetaNovas chose an approach that goes in the "opposite direction" from traditional methods. Traditional drug R&D mostly extrapolates from cells or animal models to humans, whereas MetaNovas does the reverse: starting from multi-omics data of IBD patients, it looks for metabolic pathways commonly upregulated in the patient population, then returns to animals for gene knockout validation to assess whether this pathway is involved in the occurrence and development of the disease, and further develops inhibitors against it.

AI can precisely help them shift drug discovery from "finding a needle in a haystack" to "precise navigation." XtalPi's AI + robotics platform can, in line with a project's R&D goals, rapidly identify potential candidate drugs likely to bind to specific protein targets from massive virtual molecule libraries, helping select more promising molecules for synthesis and experimental validation; MetaNovas' self-built AI platform is responsible for analyzing genomic, metabolomic and proteomic data from clinical samples.

But Xu Ke's attitude toward AI is very clear-headed: biology is not physics, not 0 and 1; there is a huge gray area in between; with the same protocol, different people in different cities may produce different results. Therefore, in his view, compared with complex biological predictions, MetaNovas at this stage values AI's application value in massive screening more.

According to the plan, MP-5342 will enter Phase I clinical trials in healthy volunteers in 2027 and Phase IIa patient clinical trials in 2028. MetaNovas is also about to complete a new round of financing to support subsequent clinical development.

The following is an excerpt from the conversation between Xu Ke, founder and CEO of MetaNovas, and Leiphone AI Technology Review, edited without changing the original meaning:

01

Entrepreneurial choice: return to China or stay?

In 2021, Xu Ke completed a joint doctoral training at Weill Cornell Medicine and Memorial Sloan Kettering Cancer Center, holding key data on the intersection of immunology and metabolism. When two American professors offered him a co-founder invitation for an immunotherapy startup, he chose a more difficult path—using immunometabolism to redefine the therapeutic possibilities for autoimmune diseases.

Xu Ke: When I left at that time, several particularly outstanding professors at our school happened to be in my field of immunology, and they had also started their own companies. Two professors invited me to join as a co-founder to develop immunology-related therapeutic drugs. That was a very good opportunity; I would not have had to work hard myself, as they had already built the operation.

But at that time I felt that what I was doing myself was more interesting—rather than choosing to be a co-founder working on someone else's thing, why not do my own thing? At that time I already had a very firm idea of starting a business.

Xu Ke: This project has always been what I was researching before, so I invited Dr. Xianyu Anjin to join.

He was already working at the time, at PatientsLikeMe, the largest patient medical data platform in the United States, and had reached the director level; the salary and benefits there were very high. But it did not take much effort to persuade him. We are all in scientific research; I did not need to persuade him with sentiment or anything else. I just said: this is my data, take a look. I did not need to say much; you will know what a great thing this is and how much potential it has. As scientific researchers, speaking with data is the most persuasive.

Xu Ke: Actually, at that time I did not have a particularly clear preference for which city to locate in—Shanghai, Suzhou, Shenzhen, all were fine. It just happened that the main investor and business partner XtalPi was here, and there would be many cooperation opportunities in the future, so we settled in Shenzhen. When we landed, we started working directly inside XtalPi.

Xu Ke: The earliest was an early-stage technology fund called IMO Ventures. They saw the articles we published and thought the immunometabolism direction we were working on was very good. It happened that IMO was also an investor in XtalPi, and they thought that since we were doing small-molecule drugs, we should be complementary to XtalPi in AI drug discovery, so they introduced us to Mr. Shuhuang. Before returning to China, we had already put the setup together.

Xu Ke: For me it was okay. Moving from scientist to entrepreneur requires gradually becoming familiar with team management, financing and product development beyond scientific research; there were indeed quite a few new topics. I was also at a stage of exploring my career direction at the time and was willing to try different ways of working, so it was relatively easy for me to enter a state of learning by doing and constantly adjusting.

▎AI Technology Review: Since becoming CEO, have there been some changes in your views?

Xu Ke: What I have always thought about is how to make the cake bigger. I hope to create value through long-term R&D and team cooperation, and I also value long-term consensus with investors. This is a relatively strong takeaway from my entrepreneurship.

Xu Ke: My own understanding is that we are a pipeline company—the so-called platforms all serve our development of pipelines. There is a platform aspect, but my positioning is the pipeline, mainly focused on developing products.

▎AI Technology Review: In the next 12 months, what is the one goal MetaNovas most wants to achieve?

Xu Ke: It is to advance MP-5342 into clinical trials and complete a new round of financing to support subsequent clinical development—drug R&D requires continuous investment.

02

Technical path: making bad cells "go hungry"

MetaNovas' core technology is built on the immunometabolism research Xu Ke has accumulated since 2013. Unlike the traditional approaches of immunosuppression or cell depletion, MetaNovas proposes a "metabolic checkpoint" mechanism: before immune cells that attack the body's own tissues "turn bad," make them "go hungry," thereby restoring immune homeostasis without causing immunodeficiency.

Xu Ke: There are many triggers for autoimmune diseases, but immune dysregulation is the common core link: IBD is mainly driven by abnormal T cell (and innate immune) responses against gut microbiota, while lupus erythematosus is prominently characterized by excessive activation of B cells assisted by T cells and the production of pathogenic autoantibodies; both involve excessive immune dysfunction that fails to distinguish self from non-self.

These immune cells share a major common feature: before they become bad cells that drive disease progression, there is a process of metabolic reprogramming—to become bad, they first have to change their metabolic foundation. You can imagine that before a bad person does bad things, he has to eat his fill; without strength, how can he do bad things? Before going to war, provisions go first. This metabolism is especially reflected in glucose metabolism; bad cells take up large amounts of glucose to obtain energy.

Our strategy is: since you must eat your fill before you can do bad things, then I will not let you eat your fill—through metabolic enzyme inhibitors, suppress the activity of its metabolic energy pathway, similar to a kind of "starvation therapy."

Xu Ke: No. The immune system is always in a dynamic balance. Immune overactivation leads to autoimmunity, but if it is suppressed too heavily, it resembles immunodeficiency, making the body highly susceptible to bacterial, viral, and fungal infections, with very serious infection risks. It must be in a balanced state—neither overactive nor too low.

Our therapeutic approach differs from existing drugs. Drugs like adalimumab (Humira) and JAK inhibitors abruptly suppress immune capacity entirely, which is why they carry boxed warnings, serious infection risks, and cardiovascular risks. We aim to return the immune state to homeostasis and balance, without needing to suppress it very low.

Xu Ke: This is also what investors ask about most: "You interfere with glucose metabolism—won't the risk be very high?" In fact, we have done extensive experiments proving that this pathway is a stress-state pathway, and cells do not use it much under normal conditions—normally cells rely on aerobic respiration, with glucose entering mitochondria to produce large amounts of energy; the glycolysis pathway is needed under anaerobic conditions and during intense exercise, and is not much needed under normal homeostasis. Therefore, inhibiting it has relatively limited impact on the function and survival of normal cells, while it shows good inhibitory effects on immune cells in a pathological state.

▎AI Tech Review: Another approach in the autoimmune field is CAR-T. How is your approach different?

Xu Ke: Some CAR-T therapies used in autoimmune disease research attempt to reset the immune system by clearing cell populations such as B cells that express specific antigens. Such treatments and their preconditioning may bring risks such as infection and require strict management. Our direction is through metabolic regulation, not clearance. The immune cells in autoimmune diseases are themselves still normal cells; they just recognize the wrong target. We just need to bring them back to balance.

▎AI Tech Review: Why choose IBD first? There should be more patients with psoriasis or dermatitis.

Xu Ke: Drug development must be related to clinical needs—you cannot look only at the number of patients, but more importantly at the efficacy of existing therapies and unmet clinical needs. First-line therapies for psoriasis and atopic dermatitis are already relatively mature and effective, so the clinical need is relatively smaller. Therefore, we prefer to concentrate limited R&D resources in disease areas with more urgent clinical needs.

For IBD (inflammatory bowel disease), the complete remission rate of existing drugs is less than 40%, and the clinical need is very large. It mainly has two types: ulcerative colitis and Crohn's disease. A slightly more severe patient may need to go to the toilet 20 times a day, looking for a toilet anytime and anywhere, afraid to go out, with constant diarrhea; severe cases can develop intestinal perforation and gastrointestinal bleeding. The cause is autoimmunity—your immune system attacks your intestines, causing long-term chronic inflammation.

Moreover, IBD patients need repeated adjustments to their treatment regimens, because the complete clinical remission rates of currently approved drugs remain limited, and some drugs target only a certain patient subgroup, making it difficult to meet the treatment needs of all patients. Therefore, a drug that is both safe and effective is greatly needed.

Xu Ke: We have tested atopic dermatitis, multiple sclerosis, primary biliary cholangitis, psoriasis, and rheumatoid arthritis in animal models, and the effects were all very good. Multiple sclerosis is when the immune system attacks neurons, leading to progressive systemic paralysis; the mechanism is very similar to ours—the only difference is the organ being attacked. There are now more than 100 autoimmune diseases, with several for each organ. Among these, the greatest clinical need is IBD; primary biliary cholangitis and multiple sclerosis also still have unmet treatment needs.

Xu Ke: Yes, and there is another pipeline with the same target but a different molecule, for primary hyperoxaluria, a rare disease that is common in children and very fatal. These children have congenital genetic defects, and the liver produces large amounts of oxalic acid; oxalic acid combines with calcium ions to form calcium oxalate, which is kidney stones. Accumulation begins from the moment of birth, and infants just a few months old can develop kidney stones. Currently there is no specific drug in China, treatment options remain limited, and some patients need to rely on hydration, pharmacological intervention, and stone management, leaving substantial unmet clinical needs.

For patients with severe disease, kidney failure may develop in childhood or adolescence, requiring organ transplantation. Because some types of the disease originate from abnormal liver metabolism, some patients need combined liver-kidney transplantation; a kidney transplant alone cannot solve the problem. This disease appears unrelated to immunity, but it happens that our metabolic enzyme can inhibit oxalate production—it is not only a node catalyzing glucose metabolism but also a node catalyzing oxalate production, a multifunctional metabolic enzyme. Therefore, the same target can be developed for indications in two directions.

Xu Ke: There are still many targets worth exploring in the metabolic network, especially for drug development targeting immune cells; many potentials are not yet recognized. It is not that technology does not allow it, but rather that people are relatively conservative about the metabolic system, thinking metabolism is important and that intervention carries safety risks; moreover, immunometabolism, as an interdisciplinary field, is still in its early stages, and not many people have noticed it.

When we returned to China in 2021, this track was blank domestically and not crowded globally either. The most direct peer is Sitryx in Oxford, UK, which also works on various metabolic enzymes, including metabolic enzymes in the same pathway but at different upstream and downstream positions; they are also a company founded by a well-known academic leader in the immunometabolism field. With the track so blank, in the early days of the startup, external presentations were basically market education and investor education.

Xu Ke: It's possible, but oncology is harder. Tumor cells have extremely strong gene mutation capacity and can quickly alter their metabolism under survival pressure—switching away from glucose metabolism to using amino acids and fats—so your drug may not keep pace with the speed of its mutations. Autoimmune disease doesn't have this problem, because the immune cells causing the disease are still essentially normal cells without extreme gene mutations; the only issue is that they recognize the host's own healthy tissue rather than bacteria or viruses.

▎AI Technology Review: What is the current clinical progress of MP-5342?

Xu Ke: We plan to enter Phase I clinical trials in 2027, mainly to assess safety in healthy volunteers. We expect to enter Phase II in 2028 to test safety and preliminary efficacy in patients.

03

Amid the wave of AI drug discovery pursuing "predicting everything," META's chosen path is to start from real patients' multi-omics data, validate in reverse, and then develop drugs. AI is applied in pragmatic places—XtalPi handles massive-scale molecule design and synthesis, while META has built its own platform to analyze clinical samples, forming a tightly interactive chain of research result analysis and validation. As for the concepts of "virtual patients" and "virtual cells," Xu Ke also maintains a rare calm.

Xu Ke: The traditional approach extrapolates from cells and animals to humans. We do the reverse—we start from patients. We analyze patients' genomics, metabolomics, and proteomics; for example, we look for metabolic pathways commonly upregulated in 100 IBD patients to see whether they represent drivers of the disease. Then we go back to animals for gene knockout validation—if we can prevent disease progression, we then develop inhibitors against that driver. Our drugs are ultimately meant to treat patients, not merely to achieve good experimental results in animal models.

Xu Ke: There are two parts. One is small-molecule R&D; we collaborate with XtalPi to combine AI-assisted molecular screening and design with compound synthesis, testing, and optimization. Small-molecule drug R&D requires exploring an extremely vast chemical space. The virtual molecule libraries used for computational screening are enormous—some reach the order of 10 to the 9th power or even larger—and it is impossible for us to synthesize and experimentally test them one by one, so we need efficient computational methods to prioritize candidate molecules worth validating.

XtalPi's AI and computational platform can predict and comprehensively evaluate candidate molecules and their drug-like properties in line with the project's R&D goals, gradually narrowing the screening scope and helping us select more promising molecules for synthesis and experimental validation. Evaluating and prioritizing such massive numbers of candidate molecules is difficult to accomplish by manpower alone, and this is also an important area where AI can deliver value in drug discovery.

In our collaboration, META mainly conducts LDHA target research and validation around immunometabolic mechanisms, while XtalPi, relying on its AI computing platform and high-throughput robotic experimental platform, is deeply involved in compound screening, synthesis, testing, and continuous optimization. The selected candidate molecules still need to be validated in wet labs; the data obtained is then fed back to train the models, continuously updating predictions and guiding the next round of molecular design and optimization. This R&D approach of mutual validation and continuous iteration between computation and experiment is, in my view, an important path for AI to deliver practical value in drug discovery.

Xu Ke: Right, the other part is our self-built clinical multi-omics patient sample analysis platform, which Dr. Xianyu Anjin is responsible for building—he previously worked on patient data platforms. This part conducts multidimensional analysis of patient samples—genomics, metabolomics, proteomics, and so on—to find patterns, validate targets, or discover new drug targets.

Why was the failure rate of drug R&D so high in the past? An important reason is that early research often relied on cell and animal models and failed to fully account for the complexity of human disease. What we focus on is a common disease feature among IBD patients: is this feature the cause of these people having IBD? Working backward, we identify the common pathway or target, then return to animals for validation—knocking out the metabolic enzyme in that pathway to see whether disease can be prevented; when disease develops, using our small molecule to inhibit it to verify whether it exerts a therapeutic effect. If it does, that indicates we have found the driver of the disease.

Starting from patients brings us closer to the real disease mechanisms and clinical needs than starting from cells or animals.

Xu Ke: Because drug R&D has a major concern: there is a huge gap between preclinical animal models and human clinical trials—efficacy seen in animals may not be seen in humans. How to reduce this gap? One approach is to use tissue samples from patients or healthy people to make organoids or organ-on-a-chip, and before entering human clinical trials, conduct preliminary safety and efficacy validation through simplified human-derived models.

In our work on autoimmune disease, we found that compared with oncology, organoid or organ-on-a-chip modeling related to autoimmune disease on the market is still at a relatively early stage of development. Relevant models in oncology are relatively mature, but immune-related ones are more challenging, because they require organoids plus the immune system to jointly form a more complex co-culture system. Our immunology background is strong, and Yaosu is very strong in organ-on-a-chip technology. If we can co-culture multiple organoids with the immune system to conduct preclinical efficacy and safety validation for autoimmune disease, that would currently be rare in China.

Yaosu is right next door to us. I have a good relationship with Dr. Xie (Xie Xin, CEO of Yaosu Technology); we knew each other before they came to Shenzhen—we are friends in the research community. It just so happens that they work in this area and there is market demand, so a conversation was enough to start a collaboration.

Xu Ke: AI drug discovery will certainly achieve a qualitative leap in the future—something like a GPT moment. But at present, biology is not something that can be quantified like mathematics or physics. Biology has too many complex variables and uncertainties. With the same protocol, two people may get opposite results. AI needs deterministic input, but the data quality of biological experiments is hard to guarantee—you may not even know where it changed or why it changed. So I believe that at this stage, compared with complex biological prediction, we value AI's application value in massive screening more.

Xu Ke: You can see the heat from the amount of financing. But this is an inevitable process: there is a hype, then it comes down and enters a cooling-off period, eliminating those with weaker technology, and then slowly a second growth curve emerges—the second wave is those who survive, have technology, and can actually produce something. This is a normal process; there must be a sifting of the sand.