Yahoo Finance

A New Paradigm in Drug Discovery: Breaking Time and Cost Barriers

New small molecule design workflow

In traditional small molecule drug discovery, the process of advancing a candidate to the clinical stage typically takes 2-4 years. This long timeline is largely due to the sequential approach of optimizing one property at a time, which often requires revisiting previous steps. However, with the advent of AI-driven workflows, the timeline is drastically reduced to under 1-2 years, with fewer synthesis and assay cycles. Virtual screening and de novo design are now indispensable steps in project initiation, followed by rapid iterations of molecular generation and experimental validation. These iterations occur within just a few weeks, allowing for parallel tracking of multiple properties. As a result, drug candidates progress to the lead optimization stage much more quickly and are further refined through an AI-guided design-make-test cycle. This new approach is 2-3 times faster, reducing total development costs by 50-70% as the project advances to the PCC stage.

New Biologics Design Workflow

Viva Biotech adopts an AI-driven antibody discovery workflow that takes a completely different route by starting with AI-generated sequences, rather than relying on traditional methods like animal immunization, which typically takes at least 2–3 months, or in vitro display. This new approach begins with AI-generated antibody sequences, a process that takes less than a week. Since these are data-driven and structure-driven rational designs, it greatly reduces the total number of sequences to be expressed (down from 10^{^9} in the case of phage display) and provides the foundation of a closed-loop learning. The lab validation step can be finished within a couple of weeks, and swift experimental feedback can be used to improve the model. The entire iteration takes less than a month, and the model refinement ensures a higher success rate in the next round.

Dr. Qian highlighted two additional advantages of this approach over conventional screening processes such as phage or yeast display:

(1) We learn from both the positive and negative data points;

(2) The models extract generalizable patterns to make it target-independent

Overall, the lab-in-the-loop AI-driven process achieves a 400% increase in efficiency and offers a success rate of over 85% in nominating at least one candidate with desired properties.

Three Powerful Modules Shaping a New Era of AI-Driven Drug Discovery

For the first time, Viva Biotech officially launches its own AI-Driven Drug Discovery platform, powered by three core modules like the three hallows. V-Scepter resembles the elder wand that empowers the computational modeling with fundamental rules. V-Orb reveals the underlying mechanisms of the biological system. And V-Mantle is the invisible cloak weaved at Viva to explore the hidden space of drug discovery with the help of generative AI models.

V-Scepter

V-Scepter includes pocket identification, force field parameterization, molecular docking, ADMET predictors and series of other useful tools.

V-Orb

V-Orb consists of active-learning augmented virtual screening as well as the molecular dynamics. Built on the enhanced sampling techniques, are our proprietary Free Energy Perturbation (FEP) calculation suits, for non-covalent, covalent binders, and biologics.

V-Mantle

V-Mantle encompasses large language models as the foundation for feature extraction and downstream tasks. De novo design and complex structure prediction are the key pillars that form a network of the state-of-the-art AI-driven drug discovery. Lastly, the antibody engineering platform represents a streamlined process with accuracy and automation.

Through a series of live demonstrations, the platform's powerful capabilities in accelerating drug discovery is further demonstrated. One highlight was the ADMET prediction tool, which delivered highly consistent results with the experimental data, underscoring the reliability and accuracy of the underlying models. Another case focused on enhanced Molecular Dynamics (MD) and Free Energy Perturbation (FEP), illustrating how the team overcame limitations of traditional computational frameworks. The demonstration emphasized the coupling between dynamic sampling and Quantum Mechanics (QM) to enable covalent interaction optimization. The entire process is brings together efficiency and accuracy in a seamless workflow.

In the field of antibody design, Dr. Qian placed particular emphasis on the comprehensive workflow implemented within the V-Mantle module. This workflow includes:

• Structure prediction

• Large language model as the foundation model of feature extraction and epitope prediction

• De novo design tools to generate antibody sequences from scratch

• Iterative design cycles to improve the affinity

• A comprehensive set of antibody developability index prediction tools to enhance the full drug profile

In the end, Dr. Qian emphasized that "The greatest strength of our approach is the connection we establish between discovery and development. This enables us to predict PK and PD profiles early on, significantly reducing downstream development risks and accelerating the overall antibody drug development process".

We firmly believe that the integration of data, algorithms, molecular structures, biological mechanisms, and both computational and experimental workflows is key to unlocking the full potential of AI in drug discovery. By uniting cutting-edge computational modeling with rigorous experimental validation, the company's AI platform is transforming the once-impossible into achievable breakthroughs. As a result, Viva Biotech is not only accelerating the pace of drug discovery and development but also redefining the very methodology of discovery, ushering in a new era of scientific innovation.

View original content:https://www.prnewswire.com/news-releases/viva-biotech-launches-the-ai-driven-drug-discovery-platform-transforming-new-drug-rd-logic-enabling-one-stop-innovative-drug-discovery-302456350.html

SOURCE Viva Biotech