On March 27, 2024, Insilico Medicine (Insilico Smart), an AI pharmaceutical company founded in 2014, updated its IPO prospectus to the Hong Kong Stock Exchange, one step closer to becoming the first listed AI pharmaceutical company in the Asia-Pacific region.

The outstanding performance of Insili Intelligence in the field of AI-assisted drug development was highly praised by Nvidia CEO Hwang In-hoon. As a company that has been growing with AI technology for 10 years, Insilicon Smart's recent research results published in the authoritative academic journal “Nature Biotechnology” made Hwang In-hoon even more excited.

(Click to watch the video)

The generative AI revolution spawns a new paradigm for drug development

At the recent Nvidia GTC conference, CEO Hwang In-hoon emphasized the importance of AI in the pharmaceutical field many times.

Wong In-hoon has always believed in the huge potential of AI+ healthcare. Fifteen years ago, he was interested in computer-aided drug discovery capabilities. Recently, Hwang In-hoon even boldly predicted that no one will need to learn programming in the future; human biology will be the most useful subject in the future.

In his keynote address at the Nvidia GTC conference, Hwang In-hoon pointed out that 10 years from now, AI can not only understand text and video, but also interpret proteins, genes, and brain waves. This will be the most revolutionary field of generative AI.

Traditional drug development is expensive and the cycle is lengthy. According to statistics, global pharmaceutical companies invest more than 250 billion US dollars in drug discovery every year.

Wall Street Insights and Insight Research was published in the article “The $6.1 billion bottleneck in the pharmaceutical industry: AI to solve it? As mentioned in “Insight Research,” a study released in October 2023 showed that the average cost for major pharmaceutical companies to market a drug from scratch has exceeded 6.1 billion US dollars. Previously, the market estimated this figure was generally between 26 and 2.8 billion US dollars.

Nvidia Vice President Kimberly Powell predicted at the just-concluded GTC conference that with the development of AI technology, pharmaceutical companies' R&D spending will increasingly shift to computing power and software. AI is expected to greatly reduce the cost of developing new drugs and shorten the drug marketing cycle.

A major paper in the Insilis Smart Nature sub-journal: How to use generative AI in pharmaceuticals?

In March 2024, Insilicon Intelligence published a major paper in the top academic journal “Nature Biotechnology”, comprehensively explaining the development process of INS018-055, the first potential “first in class” (first in class) TNIK inhibitor discovered and designed by generative AI to treat idiopathic pulmonary fibrosis (IPF). From the development of an artificial intelligence algorithm to the completion of phase 2 clinical trials, this drug candidate's pre-clinical trial and clinical trial data were disclosed for the first time.

The paper details the drug's discovery and optimization process:

1. Using PandaOmics, a target discovery engine under the Pharma.ai platform, a series of potential anti-fibrosis targets were nominated through in-depth feature synthesis, causal inference, and pathway reconstruction;
2. Pandaomics' NLP model analyzes massive text data to target TNIK as a novel target for the treatment of idiopathic pulmonary fibrosis (IPF);
3. The Chemistry42 platform combines more than 40 generative chemical algorithms and more than 500 pre-trained reward models to generate and optimize novel compounds according to structural drug design strategies to obtain drug candidate molecules INS018_055;
4. During the entire process, less than 80 molecules were synthesized and tested, greatly improving efficiency. INS018_055 showed good efficacy, safety and pharmacokinetic characteristics in preclinical and phase 1 trials.

From Pandaomics nominating the TNIK target to INS018-055 being nominated as a preclinical candidate, it only took 18 months, highlighting the high efficacy of AI-driven drug discovery.

Currently, the drug has been undergoing phase 2 clinical research in China and the US. This is also the first drug in the world to use generative AI to discover and design to enter phase 2 clinical trials. According to relevant data, only 2 drugs to treat idiopathic pulmonary fibrosis (IPF) are on the market in the world, and the global IPF market is expected to reach 5 billion US dollars in 2025.

The current prospectus also revealed a number of Pharma.ai upgrades to further accelerate AI drug discovery.

Key features of this update include:

1. Copilot architecture: deep implantation of large language models. Users can use natural language instructions to connect multi-platform functions to perform tasks such as target identification and molecule generation;
2. ADMET predictor: an integrated application of machine learning models to predict key absorption, distribution, metabolism, excretion and toxicity (ADMET) properties of small molecules;
3. AiChemistry: Using deep learning to predict binding free energy based on molecular structure and simulation characteristics.

With the support of Pharma.ai, Insilicon has recently successively nominated multiple innovative target preclinical candidate compounds, covering multiple indications such as anti-tumor and anti-fibrosis.

In December 2023, Insilicon Smart nominated two pre-clinical candidate compounds:

• Second-generation DGKA inhibitors: Pharma.ai enables further optimization of selectivity and safety, and is expected to be used to treat anti-PD-1/PD-L1 resistant solid tumors;
• Oral highly selective covalent FGFR2/3 inhibitors: Can be used to treat solid tumors “regardless of cancer type”, showing superior efficacy in multiple FGFR2/3-driven pharmacodynamic models and high-frequency mutation resistance models.

In February 2024, Insilicon Intelligence re-nominated pre-clinical candidate compounds using the chemistry42 molecular generation and optimization capabilities of its self-developed chemical engine:

A potentially “best-in-class” highly selective KIF18A inhibitor with a unique molecular framework; this compound selectively kills chromosomally unstable cancer cells, providing an innovative strategy for tumor treatment.

In addition to the development of new drugs, Insili Intelligence also released the PaperGPT paper interpretation engine developed based on ChatGPT-4 Turbo and the self-developed Large Language Model (LLM). Through the language dialogue function, the engine can provide professional answers to questions related to papers, so that ordinary readers can easily understand cutting-edge scientific research results.

Generative AI is already playing a unique role in many of InsiliSmart's businesses.

AI pharmaceutical companies are improving in terms of operating indicators

New drug development mainly includes three stages: drug discovery, preclinical, and clinical.

The success rate of drug discovery is relatively higher than drug development, but the overall success rate of drug discovery from target to emerging compound to lead compound optimization is only 51%.

Preclinical studies usually involve animal model studies to evaluate toxicology and other parameters, chemical synthesis, and optimization of drug formulations, and other studies required to ensure approval to begin clinical trials.

Clinical trials require continuous testing on healthy subjects and patients to determine drug safety and efficacy to ensure regulatory approval. The overall success rate was only 12.9%.

Faced with the dual challenges of high R&D investment and low return, many pharmaceutical companies are turning to AI-assisted drug discovery to reduce costs and increase efficiency.

However, judging from the financial data of the world's leading AI pharmaceutical companies, the industry is still in a rapidly growing investment period, but there is already a positive trend.

Take Nvidia's 50 million dollar investment in Recursion as an example,

• Since its establishment, the company has focused on building advantages through data accumulation. At present, it has more than 25PB of pharmaceutical biochemical data, which has become its core asset;
• In an automated warehouse, Recursion conducts millions of experiments every week, constantly enriching the database;
• CEO Chris Gibson revealed that the goal is to establish a basic model describing biological and chemical interactions, which is expected to fundamentally change the entire drug development process.

Recursion currently has the largest market capitalization of US$4.388 billion, but judging from revenue and cash flow, the company is still in the investment period, but the revenue growth trend is already showing.

Insight Research was published in “The $6.1 billion bottleneck in the pharmaceutical industry: AI to solve it? According to “Insight Research”, the AI pharmaceutical industry has three different business models:

1. AI SaaS: Provides AI pharmaceutical research platforms and private deployment of software to enable customer data utilization;
2. AI CRO: directly provides drug discovery results based on AI model output based on own data;
3. AI biotech: Independently develops new drugs using our own data and AI technology.

Similar to Recursion, Insili Intelligence relies on a closed loop business model from software licensing, collaborative R&D to pipeline licensing, and has reached an inflection point in revenue and cash flow.

Currently, the company's Pharma.ai localization SaaS software subscription fee is up to $500,000 per year. Foreign drug licensing has also shown potential for development. Subsequent companies are expected to continue to increase their hematopoietic capacity with more milestone gains.

According to financial data, Insili Intelligence achieved revenue of US$51.18 million in 2023, a significant increase of 70% over the previous year. Of this, $39 million comes from external licensing for drug development projects, which is the main source of revenue.

Compared with 2022, Insili Smart's operating indicators have improved significantly:

• Net loss decreased by approximately $10.19 million;
• The adjusted loss narrowed by approximately $3.44 million;
• Operating cash flow reached US$92 million, achieving a significant inflow;
• Adequate cash reserves, amounting to US$177 million;
• Net cash consumption for the fiscal year was $30.7 million, lower than the previous year.

The rise of AI pharmaceuticals reflects that global new drug development is being reshaped by the AI revolution.

Promoting drug marketing is the next milestone for AI pharmaceuticals

Although AI pharmaceutical companies are developing rapidly, the real threshold is still market verification. However, that day is getting closer.

Insili Smart's fastest progressing product, ISM001-055 mentioned above, is an innovative drug used to treat idiopathic pulmonary fibrosis (IPF), and has entered Phase 2 clinical trials. It is also the fastest drug in the world to discover and design with generative AI.

In June 2023, the first batch of administration of the drug was completed in IPF patients. This randomized, double-blind, placebo-controlled study aims to evaluate its safety, tolerability, pharmacokinetics, and initial efficacy. It is planned to be carried out simultaneously at nearly 40 research centers in China and the US.

At the same time, Insili Smart also reached two license out agreements, totaling more than 1.5 billion US dollars

Reached exclusive global license for USP1 inhibitors with Exelixis (2023.9)

• The initial payment of the partnership was 80 million US dollars, pioneering the external licensing of Insili Smart AI drugs
• Insilicon Intelligence grants Exelixis an exclusive global license to develop and commercialize ISM3091 and other USP1 targeted compounds
• In the future, it is expected to receive payment for subsequent clinical development, commercialization and sales milestones, as well as product sales commissions

Major collaboration with Menarini Group on KAT6 inhibitors (2024.1)

• Cooperation totaling more than US$500 million, granting exclusive global development and commercialization rights for menarini's novel KAT6 inhibitor
• The novel KAT6 inhibitor is expected to be used to treat ER+/HER2- breast cancer (accounting for about 70% of breast cancer patients) and other cancers. Preclinical studies have shown excellent efficacy and good safety

In addition, a number of self-developed new drugs are progressing smoothly, and two more products have entered Phase 1 clinical trials and expanded more indications

1. The QPCTL small molecule inhibitor developed in collaboration with Fosun Pharmaceutical was approved for clinical use and reached its first clinical milestone. (2023.8)
2. An oral PHD inhibitor initiated a phase 1 clinical trial to treat inflammatory bowel disease. This enter-restrictive oral small molecule inhibitor has a novel molecular framework and unique binding pattern, and has shown good safety and remarkable anti-colitis efficacy in preclinical studies. (2023.11)

AI is reshaping the global new drug development pattern, and the future of AI pharmaceuticals has arrived. As Hwang In-hoon said, human biology will become the most important subject in the future. AI will drastically reduce the cost of developing new drugs and accelerate the introduction of innovative drugs.

At the same time, as more AI pharmaceutical companies enter the capital market, the market is expected to obtain richer and more multi-dimensional business data. This will greatly broaden the boundaries of industry perception and provide investors with a clearer understanding of AI pharmaceutical applications.