news

the phase iia clinical trial stage, known as the "valley of death" in the field of ai (artificial intelligence) pharmaceuticals, has welcomed the first player to successfully break through.

on september 18, insilicon intelligent announced that the company’s pipeline ism001-055 had achieved positive preliminary results in a phase iia clinical trial. the drug is a "world's first" small molecule inhibitor driven by generative ai in the drug discovery and design process, targeting tnik (traf2/nck interacting kinase) for the treatment of idiopathic pulmonary fibrosis (ipf).

this is good news for the ai ​​pharmaceutical industry, which has frequent rumors of layoffs and integrations. however, there are currently no ai drugs in the world that have entered the clinical phase iii stage, and the business model still needs to be explored. recently, ren feng, co-ceo, chief scientific officer and head of drug research and development of yingsi intelligent, said in an interview with a reporter from "daily economic news" that the era of ai pharmaceuticals relying solely on "storytelling" to attract funds has passed. in the future, the fight is about pipeline progress and licensing income.

AIshortened early drug development time by 2/3 and saved costs by 9/10

counting from 2013, the ai ​​pharmaceutical industry has been developing for 10 years, but few ai drugs have survived.phase iiclinical trials. ai pharmaceutical star companies such as exscientia, benevolent ai, and recursion pharmaceuticals have all been involved inphase iiorⅡin the phase a clinical trial, it failed to achieve the expected drug efficacy target and experienced stock price fluctuations.

therefore, as the world’s first ai-driven drug developed and completed phase iia clinical efficacy verification, ism001-055 was highly praised by michael levitt, the 2013 nobel prize winner in chemistry: “it represents the true future of this new era of ai-driven drug discovery. breakthrough."

according to ren feng, ipf is a chronic, scarring lung disease characterized by progressive and irreversible decline in lung function. current anti-fibrotic drugs can slow down the progression of the disease but cannot stop or reverse the disease process. clinically, it is generally believed that drug administration it takes 6 months to more than a year to see the effect of the drug, but in the 3-month phase iia study of ism001-055, patients showed "dose-related improvements in forced vital capacity (fvc)" (fvc is a measure of ipf drugs) the gold standard for efficacy), proving that ism001-055 has the potential to change the disease process of ipd.

this result was beyond the expectations of the research team and needs further verification due to the short study time and small number of subjects. however, there is no doubt that generative ai can enhance the efficiency and innovation of early drug development.

ren feng said that ism001-055 took 18 months to develop from early target discovery to identification of preclinical candidate compounds, with a total r&d investment of us$2.6 million. according to relevant literature, according to the industry average, it usually takes four and a half years to complete a similar r&d process, and requires an investment of tens of millions of dollars. in other words, the addition of ai shortens r&d time by 2/3, reduces r&d expenses to 1/10 of the industry average, greatly improves r&d efficiency, and reduces r&d costs.

in addition, generative ai plays an important role in drug innovation. "first use generative ai to find new targets, and then use generative ai to target the target structure to generate new small molecules from scratch. this kind of small molecules is very innovative, and we are currently the only one in the world." ren feng said.

the reporter noticed that there are currently 18 pipelines under development disclosed on the official website of yingsi intelligent. the pipelines under development cover a wide range of indications, which is obviously different from biotechnology companies that focus on specific disease areas. in this regard, ren feng said that ai-driven new drug research and development projects are not restricted by disease fields. as long as there is sufficient public data, drug research and development can be empowered to varying degrees. however, this process has high requirements for data cleaning. the company's the data cleaning team maintains a size of 20 to 40 people all year round.

pipeline progress and licensing revenue are important. explore business models for licensing transfers.

on march 27 this year, insilicon intelligent submitted a prospectus to the hong kong stock exchange and planned to list on the hong kong main board ipo. the prospectus shows that the company's revenue consists of two parts: "drug discovery and pipeline development services" and "software solution services". the former will account for as much as 93.4% of revenue in 2023, and its subsidiary "pipeline drug development" will achieve revenue of 3902.2 million us dollars, accounting for 76.2% of total revenue.

the revenue from this important business includes research and development, commercialization after receiving marketing authorization for pipeline drug candidates developed internally by the company, and the external licensing of certain pipeline drug candidates for which the company retains exclusive ownership of relevant intellectual property rights.

based on this, ren feng summarized yingsi intelligent’s three major business models: first, external licensing of ai platform commercial software; second, external licensing of drug research and development projects; third, strategic cooperation with large pharmaceutical companies, starting from down payment, milestone payment waiting for income.

ren feng believes that among the three business models, the income "ceiling" of software licensing is relatively low, strategic cooperation is subject to the partner's strategy and is relatively passive, while drug research and development projects have higher autonomy in external licensing, and the pipeline can be completed at a certain stage. it can be transferred externally and receive down payment, milestone payment and sales share. this is not only the company's main business model, but also the business model most likely to prevail in the ai ​​pharmaceutical industry. at present, most of those willing to pay for these ai drugs are overseas customers.

"the era of storytelling for ai pharmaceuticals has passed. now everyone's requirements for ai pharmaceutical companies are twofold: one is to look at the progress of the pipeline, and the other is to look at licensing and licensing income." ren feng told reporters, with the ai ​​​​that has been listed around the world compared with pharmaceutical companies, the company's annual revenue of more than 50 million us dollars ranks second. the revenue in the first half of this year has exceeded that of last year. however, the business model has not yet been fully developed. it is "on the road to success" and needs to be updated. multiple verifications. therefore, the company's two major development directions in the future are to rapidly advance the pipeline and expand and increase revenue.

"the ai ​​platform focuses on early-stage drug development. later large-scale clinical and commercialization are not the company's strengths. therefore, when the pipeline advances to a certain stage, we must find partners." ren feng predicts that if an ai pharmaceutical company can transfer 2 ~3 projects, the entire business model can be "sustainably" implemented. for yingsi intelligent, it is expected to reach breakeven within the next 2 years.

ai is expected to shorten the gap between domestic and foreign innovative drug research and development, but it is not a “magic” for pharmaceuticals

in april 2023, the british ai pharmaceutical company benevolent ai announced that its phase iia clinical trial of the topical pan-trk inhibitor ben-2293 for the treatment of atopic dermatitis did not meet the secondary efficacy endpoint. since then, the company has conducted two layoffs. a total of 180 employees; in august this year, the american ai pharmaceutical company recursion and the british ai pharmaceutical company exscientia announced that the two parties had reached a final merger agreement, and the merged company will be named recursion.

as more and more ai pharmaceutical companies report news of layoffs and mergers, the halo that previously shrouded the industry has begun to fade, and market investment sentiment has returned to calmness from fanaticism. according to ren feng’s observation, the industry is in a stage of survival of the fittest. in the future, some companies will merge, some companies will not be able to survive, some companies will choose to transform, and some will continue to work hard in the field of ai pharmaceuticals.

so, how to judge whether an ai pharmaceutical company is good or bad? ren feng said that pipeline and revenue are two explicit indicators. the former reflects the ai ​​platform's ability to promote projects, and the latter reflects the industry's recognition of the enterprise. in addition, the evaluation of ai technology is also important, but it is more complex and difficult to judge.

ren feng believes that compared with original innovation "from 0 to 1", chinese pharmaceutical companies are better at following innovation "from 1 to 100", so they have an advantage in more engineering pharmaceutical fields such as adc (antibody drug conjugate) , is relatively lagging behind in the discovery of new drug targets such as pd-1 and glp-1. ai is expected to make up for the original shortcomings of chinese pharmaceutical companies and narrow the gap in the research and development of innovative drugs at home and abroad.

as ai pharmaceuticals empower drug research and development has become a consensus in the industry, leading pharmaceutical companies have continued to make layout moves. for example, the multinational pharmaceutical company sanofi announced "all in" artificial intelligence and data science in june 2023, glaxosmithkline, johnson & johnson , pfizer, bayer and other multinational pharmaceutical companies have integrated ai technology into the drug discovery and development process through cooperation with ai pharmaceutical startups, internal research and development, or investment. however, so far, no new drugs developed by ai-driven development have been approved for marketing in the world.

"don't think that ai (pharmaceuticals) will either fail or succeed. this is a paradox." ren feng told reporters that without ai support, the success rate of new drug development pipelines is usually less than 5%. the introduction of ai can increasing this number by 3 to 5 times is a huge improvement in itself, but it will never increase the success rate to 100%.

"everyone must have reasonable expectations for ai pharmaceuticals. it does not mean that with the blessing of ai, everything must succeed. most of them may still fail." ren feng said.

daily economic news