GENE ONLINE|News &
AI-Powered Drug Development: How NVIDIA Builds a Balance Between Biotech and TechBio
The BPIPO, NVIDIA, DCB, and TITA hosted a seminar on AI-powered drug development and precision medicine, and navigating the gap between biotech and techbio. Image: GeneOnline.
The Biotechnology and Pharmaceutical Industries Promotion Office (BPIPO), Development Center for Biotechnology (DCB), NVIDIA, and Taiwan Industrial Technology Association (TITA) collaborated to organize the “Symposium on Generative AI for Drug Development and Precision Medicine”. Held on December 19 at Taipei Bioinnovation Park, the event was co-organized by GeneOnline and attracted over 1,000 registrations, both in-person and online. The symposium brought together more than 300 guests from the biotech, technology, medical, and academic sectors, along with institutional investors, to participate at the venue.
The organizers hosted multiple presentations from NVIDIA experts, along with veteran researchers and industry elites, totaling 16 speakers. They explored the breakthroughs of generative AI (gen-AI) in the biotech and pharmaceutical industries, uncovered the immense possibilities of AI in new drug discovery and precision medicine, and provided momentum for Taiwan’s biotech innovations. As one of the co-organizers of the symposium, GeneOnline is pleased to offer the first-hand coverage below, summarizing the highlights of the event for readers to appreciate the amazing potential of gen-AI.
Related article: NVIDIA: The Rocket Fuel Boosting BIODND to Build the Most Advanced Integrated Data Platform in Biotech
Experts Expressed Hope for Accelerated AI-ization of Taiwan’s Biopharma Industry
In his opening remarks, Dr. Shiing-jer Twu, President of TITA and Chairman of DCB, mentioned that Taiwan is already the world’s leader in AI industrialization, but still needs improvement in “industrial AI-ization”, with software development ranking 26th globally, trailing its Asia-Pacific neighbors such as Korea (6th) and Singapore (3rd). He expressed his eagerness to witness the AI transformation in biotech and pharma sectors, further integrating AI with the drug discovery industry for cost reduction and efficiency enhancement.
Dr. Ettikan Kandasamy Karuppiah, NVIDIA’s Director of Technologist (Asia Pacific South), emphasized that AI technology holds great potential for healthcare and pharma applications in three key areas: digital medicine (e.g., image analysis and robotic surgery), optimization of patient-doctor interactions and digital biology, enabling more accurate and efficient medical solutions, reducing the burden of care and accelerating drug discovery. He added that NVIDIA will stay committed to these frontiers, leading the way in revolutionizing the biopharma industry for the benefit of patients around the world.
Dr. Michael Huang, Director of BPIPO and Vice President of DCB, kicked off the symposium by expressing his confidence that AI can bring revolutionary breakthroughs in Taiwan’s biotech sector, and that NVIDIA is capable of spearheading the AI transformation of the industry. He also stated that DCB will keep working with technology companies around three key aspects. The first is to promote international cooperation, intending to turn Taiwan into an ideal site for proof of concept projects by global tech and biopharma companies, and even attract them to launch IPOs in Taiwan. The second is to strengthen support for start-ups, accelerating their growth with the aegis of NVIDIA’s Inception Program. Thirdly, DCB is committed to promoting technology succession through talent cultivation programs at universities across Taiwan.
Dr. Michael Huang, Director of BPIPO and Vice President of DCB (left); Dr. Shiing-jer Twu, President of TITA and Chairman of DCB (right). (Image: GeneOnline)
Cross-Domain Integration: NVIDIA Advances Drug R&D and Precision Medicine with AI
The first session of the symposium covered insights into the trends in the AI drug development industry. Weber Liu, Director General of DCB’s Division of Industrial Development, pointed out that the number of AI drug development companies and investors worldwide has been growing rapidly over the past few years, forming a vibrant ecosystem.
In terms of partnerships and M&A, Liu mentioned that the number of collaborations between big pharma and AI tech companies has been on the rise every year, with cross-domain integration and M&A between “Techbio” and “Biotech” being particularly eye-catching. He also cited some successful examples of overseas companies to illustrate that many of these M&A deals are about combining the strengths of both parties, creating synergies by leveraging on each other’s unique assets.
Dr. Inca Chen, NVIDIA’s Healthcare & Life Sciences Solutions Architect, stated that the drug discovery industry has reached a pivotal turning point. AI breakthroughs now allow researchers to quickly solve protein structure puzzles, process vast amounts of gene sequencing data, and sift through thousands of compounds to identify the most promising ones for further testing. She highlighted the biological applications of large language models (LLM), noting that biomacromolecules like DNA, RNA, and proteins have specific sequences. LLM can process and decode these sequences, similar to how ChatGPT analyzes human language.
Dr. Chen showcased success stories of how NVIDIA’s AI tools are driving breakthroughs. For example, Amgen used NVIDIA’s computing power to build gen-AI models for antibody sequence design and molecule screening, speeding up the development of new antibody drugs. Receptor.AI, a UK-based company, integrated NVIDIA BioNeMo cloud APIs with its end-to-end computer-assisted drug discovery (CADD) platform. This integration accelerated virtual drug screening, ADMET assessment, and protein-ligand docking prediction, improving drug discovery outcomes and reducing costs.
In gene therapy and precision medicine, Dr. Chen discussed how NVIDIA GPUs and the BioNeMo framework helped Dyno Therapeutics train its protein language model. This model improved the capsid protein of adeno-associated viruses (AAVs), enhancing gene delivery to specific cells. Additionally, a Stanford Medicine team used NVIDIA’s GPU-powered solutions for accelerated genomics. They sequenced a patient’s genome in about five hours, setting a Guinness World Record. This achievement highlights how AI is advancing precision medicine and genomics.
From left: Weber Liu, Director General of the Industry Development Division, DCB; Dr. Inca Chen, Healthcare & Life Sciences Solutions Architect at NVIDIA; Dr. Jer-Wei Chang, Deputy Leader of the Intelligent Medicine Division, DCB; and Dr. William Wong, Senior Developer Relations Manager at NVIDIA. (Image: GeneOnline)
Optimizing Drug Target Exploration with Gen-AI
The second session of the event focused on “AI-powered Drug Discovery and Target Analysis”. Dr. Jer-Wei Chang, Deputy Director of DCB’s Intelligent Medicine Division, reiterated the three major dilemmas facing the drug development industry: high cost, long timeframe (often taking more than 10 years), and low success rate (over 80% failure rate). The key to overcoming these challenges lies in the cross-domain integration of technology and life sciences through applications of big data and AI.
Dr. Chang recalled that back in 2002, some scientists already suggested that about 10-15% of the genes in the human genome are druggable (i.e., suitable for use as targets of new small molecule drugs). In recent years, the definition of “druggability” has been expanded to include new drugs such as proteins and antibodies, and even novel agents such as proteolysis-targeting chimeras (PROTACs), molecular glue, and exosomes.
Through AI tools and machine learning, researchers can combine biological data such as genomics, proteomics, and drug response to rapidly predict potential targets and design new ligands, as well as discover new druggable targets, expediting the process of drug discovery and screening for candidates.
Enhancing Drug Target Discovery and Multi-Target Cancer Therapy Development Using Gen-AI and Causal AI Approaches
Dr. Yu-Feng Wei, Co-founder and CEO of Vizuro, introduced the concept of Causal AI and its biomedical applications. Causal AI identifies and interprets causal relationships in data, analyzing the actual impact of one variable on another through causal inferences while considering potential interfering factors. “Causal AI is still a blue ocean in the biopharma field,” said Wei, adding that any company that can create highly efficient solutions could generate tremendous value.
Dr. Wei also discussed the potential applications of Causal AI in drug development and precision medicine, including precise searches for disease targets and biochemical pathways, drug repositioning (new indications from existing drugs), drug design optimization, and the identification of biomarkers for clinical populations. Taking cancer treatment as an example, generative Causal AI helps design new drugs that can simultaneously address multiple target genes, destroying cancer cells while minimizing damage to normal cells through “synthetic lethality” (triggering cell death by inhibiting two non-lethal genes simultaneously). Additionally, Causal AI assists in designing antibodies against neoantigens, facilitating the development of cancer vaccines. In precision medicine, Causal AI applies to biomarker-driven targeted therapies, as well as personalized early detection and prevention of cancer.
Dr. Charles Chaung, Bioinformatics Scientist at Atgenomix, discussed the new opportunities for target exploration and drug discovery brought by AI applications from a multi-omics perspective. He began by addressing the challenges of precision medicine before the maturity of multi-omics techniques. Taking gastric cancer as an example, he highlighted that although it is well known that patients with the diffuse type of the disease have a worse prognosis, there was still limited understanding of its causes and treatment efficacy. Fortunately, with the advancement of gene sequencing techniques, researchers can now categorize gastric cancer into four molecular subtypes based on genomic data, enabling doctors to find the right drugs according to the underlying mutations.
Dr. Chaung explained that multi-omics analysis allows researchers to better understand the true nature of various diseases, inspiring drug discovery and development. Atgenomix, for example, utilized NVIDIA GPUs and generative AI to accelerate the analysis of multi-omics data. Using the standards of the Global Alliance for Genomics and Health (GA4GH), the company built the “Atgenomix SeqsLab” platform to rapidly analyze and manage massive amounts of biomedical data. With NVIDIA Parabricks, SeqsLab reduces the time required for whole genome sequencing (WGS) from 5 hours to just 10 minutes. Additionally, with NVIDIA RAPIDS, an open-source suite of GPU-accelerated data science and AI libraries, the platform can perform RNA sequencing in 10 minutes instead of over 2 hours, achieving both cost reduction and improved accuracy.
Moreover, with NVIDIA NIM microservices and LangChain technology, the platform can avoid generative AI hallucinations through retrieval-augmented generation (RAG). In summary, the comprehensive integration of generative AI, clinical knowledge, and multi-omics analysis connects complex multi-omics and empirical data, providing doctors with more customized treatment strategies.
Development of AI-powered Drug Design Optimization and Validation
The third session featured in-depth discussions on the development of AI-powered drug design optimization. Dr. Yi-Yu Ke, Director of DCB’s Division of Intelligent Medicine, summed up his experience in drug R&D over the past two decades and shared the evolution and application of AI in drug discovery. He explained that with the aid of generative AI, it is possible to quickly handle quantitative structure-activity relationships (QSAR) of drug molecules and complete virtual screening of hundreds of thousands of compounds in a matter of 20 minutes, exporting the most promising candidates for experimental validation. These innovations enable a cost-effective development process that would otherwise take ten years to complete in one to two years.
Dr. Ke also highlighted the use of AI in small molecule and antibody drug design, including AI-assisted pharmacokinetic analysis and molecular dynamics simulation to predict the activity and toxicity of molecules at the drug design stage. With the increasing sophistication of AI and big data, the amount of data needed to train new models will be further reduced, and the range of AI applications will expand from small molecules and antibodies to nucleic acid drugs.
Tim Chen, Director of AI Platform of Graphen Drugomics (Taiwan Branch), focused his presentation on the breakthrough applications of multimodal AI in drug discovery and development. He pointed out that multimodal AI can combine data on protein structure, chemical molecules, and medical imaging to accomplish drug screening, design, and optimization under a unified framework. Instead of generating a large number of structures and then screening them, researchers can design and generate compounds that could qualify as drug candidates based on druggable targets, leading to a new generation of drug manufacturing processes.
Further, experts shared success stories of Gen-AI-powered drug design. Hsin Liu, co-founder and CTO of VIRTUALMAN Inc., highlighted how AI optimizes drug synthesis by generating candidate compounds and predicting synthesis pathways, reducing the design-to-lab-testing time to two weeks and increasing success rates. Dr. Shu-Jen Chen, Chief Scientific Officer of AnHorn Medicines, discussed the use of AI in designing protein degraders. By leveraging generative AI and NVIDIA’s BioNeMo platform, AnHorn generated a library of 10 million compounds, rapidly screening and assessing potential drug candidates. This approach reduced development time from 4.5 years to just 14 months, cutting costs by 560-fold.
Taiwan’s AI Drug Development Still Holds Great Potential for Progress
The final session focused on Taiwan’s potential in AI-driven drug development. Dr. Yen-Chu Lin of National Yang Ming Chiao Tung University highlighted the success of ISM001-055, an AI-designed drug for idiopathic pulmonary fibrosis (IPF) that showed positive topline results in a Phase 2a clinical trial. Despite the growing prevalence of AI in the drug development industry, Dr. Lin mentioned that most studies on AI-driven drug design (AIDD) remain in clinical phases 1 or 2, while Taiwan still lingers at the stage of exploring new targets and preclinical studies. Thus, he urged biotech companies to align innovations with market demands.
Dr. Jung-Hsin Lin from Academia Sinica discussed advancements in AI protein structure prediction, highlighting the need for improvements in predicting intrinsically disordered regions (IDRs) and intrinsically disordered proteins (IDPs). He also pointed out that generative AI has yet to visualize the structure of proteins in different functional states. Additionally, determining the locations of metal ions, cofactors, and ligands remains challenging. To address these limitations, Dr. Lin’s research team is developing new AI prediction models and quantum physics-based correction methods. New AI models and tools like idTarget and SLITHER are being created to enhance drug development accuracy and efficiency.
Dr. Lee-Wei Yang of National Tsing Hua University emphasized the importance of considering absorption, distribution, metabolism, excretion, and toxicity (ADMET) factors in AI drug design. His team, in collaboration with UT Southwestern, used AI algorithms to develop a lead compound targeting EglN2, a biomarker for triple-negative breast cancer, and demonstrated significantly lower toxicity. He also cautioned that current AI tools, like AlphaFold3, don’t fully account for intermolecular forces, which could affect drug design success.
Further, Dr. Ren-Hua Chung from the National Health Research Institutes discussed polygenic risk scores (PRS), which assess an individual’s genetic risk for diseases. While PRS is growing in importance, most studies are based on white populations, leaving Asian populations underrepresented. About 80% of relevant studies are based on data from white populations. Given that whites constitute only 10-15% of the world’s population and Asians account for over 60%, Asian populations are under-represented in PRS-related studies, leaving plenty of room for future improvement. The International Consortium of Integrative Genomics Prediction (INTERVENE) is working to enhance PRS tools with biobanks from Japan, Qatar, Taiwan, and China, aiming to improve disease prediction accuracy for conditions like atrial fibrillation, breast cancer, and type 2 diabetes.
NVIDIA’s Inception Program Fuels Growth of AI Pharma Startups
Another highlight of the symposium was Dr. Sheng-Ta Lee, NVIDIA’s Head of Startup & VC Partnerships (Taiwan), introducing the NVIDIA Inception program. Participating startups can access NVIDIA’s hardware and software discounts, education and training programs, and performance enhancement services for hardware and software integration. For companies in the fundraising stage, NVIDIA also provides assistance with financing, product marketing, and participation in international exhibitions.
As of September 2024, more than 23,000 startups have joined the Inception program, connecting with over 1,000 venture partners. AI pharma startups like AnHorn Medicines and VIRTUALMAN, which participated in the symposium, have also enrolled in this program.
DCB and Vizuro Join Forces to Create AI Drug Development Platform
Immediately after the event, DCB and Vizuro LLC held a signing ceremony for their memorandum of understanding. Both parties will collaborate to establish a new-generation AI platform to drive breakthroughs in drug discovery and advance precision medicine. The agreement outlines a partnership across nine key areas, including small molecule drugs, antibody-drug conjugates (ADCs), bispecific antibodies, nucleic acid drugs, and cancer vaccines. By leveraging DCB’s wet lab facilities and compound libraries, alongside Vizuro’s causal AI technology platform, the two aim to expedite new drug discovery and repurpose existing drugs. Their goal is to bring at least one AI-designed drug candidate to the clinical stage within 18 months.
The symposium highlighted the diverse applications and immense potential of AI in drug discovery, from generative AI and causal AI to multimodal AI. It showcased innovative tools and technology platforms that are reshaping the traditional drug discovery process and overcoming the industry’s challenges of high costs, long timelines, and low success rates. With the combined efforts of BPIPO, DCB, and TITA, alongside NVIDIA’s advanced AI hardware and software packages, generative AI is set to become a key driver in new drug R&D and precision health. This collaboration will benefit the public and patients, while also helping Taiwan’s biopharma industry advance into a new era of cross-domain integration.
©www.geneonline.com All rights reserved. Collaborate with us: [email protected]
Author
LATEST
