When proteome meets AI, what is accelerated

What meaningful or interesting information can an ordinary person obtain by performing mass spectrometry detection of the whole blood proteome? Who will be the 23andMe (DNA identification company) in the proteome space?

In July last year, when 98.5% of the protein structure of the human proteome was deciphered by the AlphaFold algorithm, many netizens raised this question on Zhihu. Some investors simply ask listed companies directly whether they have development plans and technical reserves involving artificial intelligence (AI) in the field of life sciences.

Although proteomics research is not a new concept, with the breakthrough of AI technology, the development and marketization of proteomics-related applications is rapidly fermenting.

At the beginning of this year, West Lake Omi (Hangzhou) Biotechnology Co., Ltd., a provider of artificial intelligence proteome diagnostic services, announced the completion of hundreds of millions of yuan in Pre-A round of financing. The clinical laboratory self-built project (LDT) product they developed for the diagnosis of thyroid nodules is also coming out soon.

Will proteomics + AI be the next market focus?

When proteome meets AI, what is accelerated

New algorithm reshapes proteomics

Some people joke that the primary reason why the concept of proteomics is cold in the secondary market is that people do not know what it is.

But in fact, it is the key to unlocking precision medicine.

The magic drug “Gleevec” in the movie “I’m Not the God of Medicine”, its target is a fusion protein, and the fusion protein is inhibited by small drug molecules, which ultimately controls the development of chronic myeloid leukemia. In other words, whoever “understands” proteins better can find the key to cracking major diseases and developing new drugs.

However, “traditional proteomics analysis techniques and methods are not completely suitable for studying protein systems. What is lacking is the process of quantitative data accumulation for proteins, and there is no suitable algorithm.” Distinguished researcher of West Lake University, West Lake Guo Tiannan, founder of Omi, said.

The emergence of AlphaFold has brought light to the scientific community.

In July 2021, the artificial intelligence company DeepMind and the European Institute of Bioinformatics cooperated to release the protein structure database predicted by AlphaFold, which completed 98.5% of the protein structure prediction of the human proteome. This is considered one of the most important scientific breakthroughs of this century.

Subsequently, DeepMind published the source code of AlphaFold2 in “Nature”. AlphaFold2 can decipher the three-dimensional structure of ordinary proteins in minutes, and can also predict the structure of a large protein connected by 2180 amino acids.

“Before Alphafold2, AI technology has never really entered the microscopic life world. But in fact, AI technology is not only a necessary condition but also a sufficient condition for the microscopic life world.” Guo Tiannan said.

Without AI technology, human understanding would not be able to make sense of so many protein dynamics because it is simply too complex. “Genes are relatively stable. It is enough for a person to do gene sequencing once in a lifetime, but proteins change all the time. A cold will change many cells and proteins. Precision medicine is to provide the most suitable diagnosis for a patient based on his current condition. And treatment methods. With proteomics + AI, precision medicine will be raised to another level.” Guo Tiannan said.

AI powers precision medicine

More than ten years ago, Guo Tiannan studied and worked in the Department of Hematology of Wuhan Union Medical College Hospital. He clearly remembered that a box of Gleevec was more than 300,000 yuan, and the department bought it and sold it to patients one by one. Later, he went to ETH Zurich, Switzerland, under the tutelage of Ruedi Aebersold, one of the pioneers in the field of proteomics.

After returning to China in 2017, he brought high-throughput mass spectrometry to the laboratory of West Lake University. In short, this technology is to “measure the weight” of tens of thousands of proteins, and to distinguish between A and B by being accurate to 30 decimal places. You must know that in the microscopic world, proteins are always changing, and there is no technology that can identify them through molecular surface features. High-throughput mass spectrometry technology can “film” the movement of proteins, and record the big data generated by protein changes when the drug enters the cell.

“Mass spectrometry is used to record molecular weight, and high-throughput requires fast speed, so as to deduce what kind of protein it is and how much it is.” Guo Tiannan told the China Science Journal.

Precision medicine mainly involves two aspects: diagnosis and treatment, and Xihu Omi has a layout in both aspects.

Thyroid nodules are a high incidence in the population. 30% of thyroid nodules cannot be identified as malignant or benign. Many patients suffer unnecessary cuts under psychological pressure, and patients who have lost their thyroid must take medicine for life. The LDT product developed by Xihu Omi combining proteome with AI technology can make most of the 30% of patients avoid surgery.

Guo Tiannan also has a “small goal”, which is to make it an inexpensive diagnostic method through the development of LDT products whose cost is reduced to 1/10.

What excites scientists and the market even more is that AI also gives pharmaceuticals more room for imagination. Different from domestic ones, foreign pharmaceutical companies attach great importance to the development of new drug targets. At present, about 500 drug targets have been discovered, and about 40% of the drugs are discovered and designed with G protein-coupled receptors as targets.

In AI pharma, high-throughput mass spectrometry is recognized as the most efficient measurement option for practical use. In Guo Tiannan’s view, we can start at least in two aspects: generate protein data related to pharmaceuticals, combine AI models, establish new methods that are more effective than drug screening, and promote new drug research and development; establish a new technology based on protein, big data and AI technology. Drug production and quality control processes to find the best drug production method.

Currently, he is cooperating with a number of international pharmaceutical companies to develop drug targets, and cooperates with hospitals and pharmaceutical companies to carry out related pharmaceutical work.

Imagination could be bigger

Compared with the secondary market, it has not yet “showed up”. In recent years, proteomics has ushered in its “little spring” in the primary market.

In 2020, protein engineering service provider Baipu Bio announced the completion of tens of millions of Pre-A round financing; proteomics technology service and product seller Zhongke New Life announced the completion of 200 million yuan round A financing; proteomics technology development and application Shangjingjie Biotechnology announced the completion of the B round of financing of 530 million yuan.

Westlake Omi, where Guo Tiannan is located, announced on January 13 that it had completed hundreds of millions of yuan in Pre-A round of financing. This round of financing was jointly led by Yifeng Capital and Hillhouse Ventures, and followed by Gaorong Capital, Powerfang Capital and Westlake Ventures. Last year, Westlake Omi has completed the angel round and angel + round of financing.

Today, proteomic methods are used in the clinical research, diagnosis and treatment of kidney cancer, liver cancer, colorectal cancer, lung cancer, gastric cancer and other cancers. Scientists say a new era of proteomics-driven precision medicine (PDPM) has arrived.

Le Beilin, executive director of Gaorong Capital, told China Science News that companies that use proteins for diagnosis, drug antibody development and protein raw materials have been favored by capital before, and they represent the application of proteins in terminals and raw material intermediates respectively. At the same time, deploying the upstream end of the protein industry, such as the newly discovered big data end proteome, has also attracted much attention in recent years.

“In the past, proteomes were limited in sample size, type, and the amount of data generated per sample. Therefore, we prefer companies that have innovative and unique techniques in basic data methodology and can generate high-quality data. ‘ said Le Belling.

Mastering the core technology of basic methodology can realize the analysis of the proteome of very small and trace samples. He has a variety of clinical sample processing capabilities, unique data analysis and interpretation capabilities, and Guo Tiannan, who has a medical background, is a very good “translator” of the company. “… These are the reasons why West Lake Omi is favored by the capital side.

From the launch of the Human Genome Project to the last decade, a number of listed companies have emerged at home and abroad, such as BGI, GRAIL, a tumor NGS company, and SEER, a proteomics company. “Proteomics can theoretically replicate the path of the genome in diagnosis, and even go deeper. Looking further, the application of proteomics can move from diagnosis to treatment, and there may be more room for imagination.” Le Bellin said, this is also true Guo Tiannan’s confidence and expectations in trying AI pharmaceuticals at this stage lie.

Whether it is possible to step into the deep water area, discover representative drug targets, and make them into new drugs is the goal that scientists are striving for. For example, He Fuchu, an academician of the Chinese Academy of Sciences and president of the Academy of Military Medical Sciences, discovered a good target in the direction of liver cancer, and conducted drug screening on the target. “In the next step, whether the technical accumulation of proteomics can make a breakthrough in the innovative research and development of drugs, we will wait and see.” Le Bellin said.

Read more here: Source link