BioGeometry Secures $10M Series Angel Funding to Advance AI-Driven Macromolecular Drug Discovery

VCBeat learned that BioGeometry has completed a multi-million-dollar angel round of financing, exclusively invested by Gaorong Ventures. Currently, the team has also released TorchProtein, its first open-source machine learning platform for large molecule drug research and development, aiming to accelerate the drug R&D process through AI. According to Dr. Jian Tang, founder of BioGeometry, this round of financing will mainly be used for subsequent team expansion, etc.

Why choose to establish BioGeometry in 2021?

Dr. Jian Tang, Associate Professor and Tenured Professor at Mila, the Algorithm Institute of the University of Montreal in Canada, and Founder of BioGeometry, stated that, to a certain extent, it is based on the current industry's technological innovation combined with the technical accumulation of the team he leads, which has prepared the technology and talent reserves for AI-driven drug research and development.

From the perspective of industry technology development background"At present, we are at the intersection of the AI and biotechnology revolutions. On one hand, geometric deep learning technologies (such as AlphaFold2) have made significant breakthroughs in molecular modeling; on the other hand, biotechnologies represented by synthetic biology enable rapid reading, writing, and editing of genes, providing AI with vast amounts of data. The deep integration of these two revolutionary technologies brings tremendous opportunities for biomacromolecule design," mentioned Dr. Jian Tang.

AndBioGeometry's founding team happens to have deep technical accumulation in related fields.Dr. Jian Tang, the founder of BioGeometry, graduated from the School of Information Science at Peking University. He has served as a researcher at Microsoft Research Asia and conducted postdoctoral research at Carnegie Mellon University and the University of Michigan in the United States. Dr. Tang has done many pioneering works in the field of graph representation learning and is one of the few scholars internationally who first applied deep learning to graph-structured data. He received the best paper award at ICML'14, one of the top three conferences in the field of machine learning (the only one in China), and was nominated for the best paper at WWW'16, a top conference in the field of data mining.

At that time (2012), many teams were applying deep learning to fields such as computer vision and natural language processing, but Dr. Jian Tang had already been applying deep learning to graph-structured data. His representative work in the field of graph representation learning, LINE, published in 2015, has been widely recognized by universities and industries both in China and abroad, with over 4,500 citations. In 2018, he joined Mila, the Algorithm Institute at the University of Montreal in Canada. Notably, Mila was founded by Turing Award winner and one of the three giants of deep learning, Professor Yoshua Bengio. It is one of Canada's three national laboratories in the field of artificial intelligence and currently the largest academic AI laboratory in the world.

At the time, Dr. Tang Jian was searching for the "killer application" of graph neural networks. He realized that most data in the biopharmaceutical R&D field is graph-structured, meaning his algorithms had found their place. Subsequently, he led his team to apply graph neural networks to the biopharmaceutical R&D field. Initially, the team focused on small-molecule drug development, but with the success of AlphaFold 2 in molecular modeling, they also began applying graph neural networks to large-molecule drug development.

The early technical reserves created conditions, but what ultimately led them to choose to move from academic research to commercial implementation and start a company was because they sawThe AI-driven large molecule drug R&D field, still a blue ocean, has tremendous room for development.——

In the biopharmaceutical field, cell therapy and gene therapy are increasingly playing a more significant role. After witnessing the "miracles" brought by biomedicine time and again, the public, who have long suffered from diseases, place high hopes on the development of biomedicine. At present, a number of companies that apply AI to small-molecule drug research and development have emerged in China, but there are only a few AI companies stepping into the limelight in the large-molecule drug research and development field. The enormous unmet demand has led Dr. Jian Tang's team to decide to dedicate themselves to this cause. Thus, BioGeometry was established.

Perhaps, the halo of "technical accumulation" is also the reason why BioGeometry has gained favor from Gaorong Ventures.Gao Rong Ventures Founding Partner Yue BinMentioned, "Breakthroughs in the computing field are reshaping the process of drug discovery. We believe that artificial intelligence can significantly advance the development of large-molecule drugs. Dr. Jian Tang has applied graph representation learning and geometric deep learning techniques to the field of drug research and development, conducting numerous pioneering works. He has also achieved internationally leading technology in antibody optimization and antibody structure prediction tasks. We look forward to BioGeometry accelerating the drug R&D process through next-generation artificial intelligence technologies and addressing major disease challenges."

Perhaps we should not define BioGeometry as a large molecule drug development platform.

Building on prior research, Dr. Tang Jian's team has developed TorchDrug, an open-source machine learning system specifically designed for drug discovery (primarily targeting small-molecule drug development). This system aims to promote open-source sharing of AI in the field of drug discovery and accelerate advancements in drug development. It has already garnered significant attention previously.

Now, the BioGeometry team has also collaborated with companies such as NVIDIA, Intel, and IBM to release the first open-source machine learning platform for large molecule drug development, TorchProtein. The platform has open-sourced a general framework for deep learning modeling of large molecules, the first pre-trained large model based on the three-dimensional geometric structure of proteins, and a standard dataset specifically used to evaluate the effectiveness of deep learning in protein modeling.

Dr. Tang Jian mentioned that one of the unique aspects of the open-source system developed by BioGeometry is that they had already conducted extensive in-depth academic research on fundamental issues in related fields earlier on, and based on this research, they improved their algorithms. Regarding the open-source machine learning platform TorchProtein,Dr. Tang Jian also mentioned"At present, the common approach in the industry for protein pre-trained models is to borrow natural language processing analysis techniques, representing proteins as sequences to construct pre-trained models. AndWe have provided a set of groundbreaking algorithms that are pre-trained directly based on the three-dimensional geometric structure of proteins. Its advantage lies in better extraction of protein features. After extensive pre-training, the reliance on labeled data will be significantly reduced in later stages, further lowering R&D costs.”

It is reported that BioGeometry has completed the construction of the AI large-molecule drug design platform. Moreover, to achieve a closed-loop of dry and wet experiments, BioGeometry is also gradually building a high-throughput large-molecule drug wet-lab validation platform. Considering the time and cost required for the latter platform, in the initial stage, the company plans to collaborate with well-known universities and laboratories in the biopharmaceutical field to complete relevant experimental validations.

Regarding future development plans, Dr. Tang Jian introduced that in the research and development of macromolecular drugs, they are currently more focused on projects where BioGeometry has already gained certain international advantages, such as antibody structure prediction, antibody optimization, antibody sequence design, and enzyme activity prediction. Subsequently, they will further expand their focused niche areas. In addition, although BioGeometry is a startup project, it is also attempting to gradually expand from academic cooperation to business cooperation based on the relationships established earlier with international pharmaceutical companies.

In order to achieve this goal, BioGeometry is actively building a more competitive team. Currently, its team already includes interdisciplinary talents with diverse backgrounds, such as bioinformatics, structural biology, and top AI research talent. Turing Award winner, one of the three giants of deep learning, and professor at the University of Montreal, Yoshua Bengio, will also serve as a company advisor. After the completion of this round of financing, the team will also attract more talent with AI and biopharmaceutical backgrounds.

Facing the current situation of talent shortage in the AI pharmaceuticals field, how can suitable talent be found? Dr. Jian Tang appears unconcerned about this issue. "In the early stages, we established extensive collaborations with well-known university laboratories both domestically and internationally, which has provided us with some advantageous conditions for expanding our influence and acquiring talent."

But compared to building a team, the focus of future development for Dr. Jian Tang may still be the construction of a high-throughput macromolecular drug wet-lab validation platform. "Once our high-throughput macromolecular drug wet-lab validation platform achieves intelligence, datafication, and automation, it can form a closed loop with the AI-driven macromolecular drug design platform. Through this integration of dry-lab and wet-lab experiments, we can rapidly complete candidate drug design and improve the success rate of candidates in clinical stages, ultimately greatly accelerating the entire drug development process."

Gaorong Ventures, founded in 2014, focuses on early-stage and growth-stage investments, with key investments in new consumption, new technology, healthcare, and other innovative entrepreneurial fields. To date, 20 companies invested in or partnered with by Gaorong have gone public, and over 30 companies invested in or partnered with by Gaorong are valued at more than 1 billion US dollars. Several of these companies have grown into leaders in their respective industries, including Pinduoduo, Huya Live, BOSS Zhipin, Roborock, and Dingdong Maicai. Gaorong Ventures continues to focus on the healthcare sector, discovering and supporting leading enterprises in digital health, medical services, new drug development, medical devices, and testing.