How is AI transforming the development of therapeutic antibodies?
Therapeutic antibodies are a significant advancement in modern medicine. They offer hope for treating many diseases, including cancers, autoimmune disorders, and infectious diseases.
However, The path from idea to clinic is full of challenges due to the complexity of biological systems and strict approval standards.
This article outlines the stages of antibody development, explores the challenges, and highlights the potential of artificial intelligence (AI) in transforming the field.
The Stages of Antibody Development
The Role of Artificial Intelligence in Antibody Development
Artificial intelligence stands at the forefront of transforming antibody development. AI and machine learning (ML) algorithms can analyze vast datasets, uncover patterns, and predict outcomes with a speed and accuracy unattainable by human researchers. The integration of AI into the R&D process offers several promising avenues:
Accelerating Antibody Development: AI-based methodologies hold great promise for accelerating and improving the development of therapeutic antibodies. These technologies are increasingly integrated into development pipelines, overcoming the limitations of the traditional, largely empirical antibody development process.
Improving Antibody Discovery and Design: Recent advances in AI technology have had an incredible impact on antibody design and discovery, resulting in significant advances in antibody discovery, optimization, and developability. AI techniques, including traditional machine learning and deep learning, have played decisive roles in evaluating the efficacy of antibody immunotherapy and providing accurate predictions for optimizing drug treatments.
Efficiency and Cost Reduction: AI promises to increase efficiency and lower the cost of antibody discovery and development. It can transform the biopharmaceutical industry’s research and development, leading to rapid response and improved patient therapeutic access.
Custom-Built Antibodies: AI-driven approaches can be used to predict the ability of antibodies to penetrate cells and optimize critical features such as their half-life in vivo, solubility, immunogenicity, and toxicity. This approach shortens the time and costs of drug development and progression.
Rapid Identification of Effective Antibodies: AI-based antibody discovery platforms enable the rapid isolation, characterization, and identification of effective antibodies with good developability attributes. This benefits cost-of-goods and improves therapeutic access to patients. AI research tool rapidly analyze vast datasets, including patents and research papers, to identify potential antibody candidates.
However, AI also comes with limitations that need to be addressed:
Limited biological understanding: Current AI systems lack the deep, intuitive understanding of biological processes and disease mechanisms that human experts possess. This could lead to overlooking important context or nuances.
For example, An AI system may identify a novel antibody target based on data patterns, but it lacks a deeper understanding of why it is relevant to the disease biology and pathways.
Data dependency: AI models rely heavily on the quality and quantity of data they are trained on. Biases or gaps in the training data can propagate into the model’s predictions and development strategies.
For example, if an AI model’s training data lacks sufficient examples of antibodies targeting certain epitopes or mechanisms, it may overlook or inefficiently explore those areas.
Lack of creativity: AI excels at optimizing within defined parameters but may struggle with the open-ended creativity required to generate truly novel solutions or overcome unexpected roadblocks.
For example, When hitting roadblocks like manufacturing issues or off-target binding, an AI may struggle to devise the creative protein engineering strategies a human researcher could implement.
Interpretability issues: Many AI models, especially deep learning approaches, operate as “black boxes,” making it difficult to interpret how they arrive at specific predictions or design decisions.
For example, A deep learning model predicts an antibody candidate will be highly potent but cannot explain the key residue interactions or structural features driving that prediction.
Regulatory and ethical concerns: Using AI for critical healthcare applications like antibody therapies raises regulatory, ethical, and liability questions that currently lack clear guidelines.
For example, If an AI system’s decisions directly influence clinical trials or therapy selection, there are open questions about liability and patient consent.
Implementation challenges: Integrating AI systems into antibody development pipelines and infrastructures requires significant computational resources, expertise, and process re-engineering.
For example, Integrating an AI system that requires substantial GPU computing power into an existing antibody discovery pipeline could involve overhauling infrastructure.
Despite the cons, the industry now sees a ray of light in using AI to speed up and make antibody development more cost-effective.
Innovative Startups using AI to reduce the time and cost of Antibody Development
1. Denovai Biotech helps bring the early discovery and screening process down to 1 year
Development Stage Impacted: Discovery and Target Identification, Antibody Generation and Screening
Solution Impact and Time Savings: Denovai Biotech’s AI algorithms enhance the discovery process by efficiently analyzing vast datasets to identify potential targets and generate antibodies with high specificity and affinity.
Traditional methods in these stages can take up to 2 years (24 months), combining the times for target identification (6-24 months) and antibody generation and screening (12-18 months).
By streamlining data analysis and prediction processes, Denovai could potentially speed up the discovery process from months to days, reducing the early discovery and screening process to one year or less. This significant time reduction is achieved by rapidly processing complex biological data and identifying high-value targets and antibody candidates much quicker than traditional methods.
2. ImmunitoAI helps bring the preclinical development stage down to a few hours
Development Stage Impacted: Engineering and Optimization, Preclinical Development
Solution Impact and Time Savings: ImmunitoAI focuses on making the AI decision-making process more interpretable, which is crucial during the antibody engineering and optimization stages.
These stages traditionally take 6 to 24 months, depending on the complexity of the modifications required to enhance antibody properties.
ImmunitoAI can help researchers make faster and more informed decisions on which antibodies to advance to preclinical development by providing insights into why certain antibodies are predicted to be more effective.
The startup’s unique AI product designs and generates thousands of sequences that have the potential to bind to a specific target.
These sequences are masked for confidentiality purposes. The startup follows a “drug-first” approach, where the platform screens for drug properties at various stages.
The first screening stage involves analyzing the generated molecule sequence. Subsequently, the sequence is folded through protein folding prediction, which typically takes a few hours.
In 2021, ImmunitoAI raised $1 million in a seed funding round. Google also invested in it in 2023.
3. EVQLV to make the screening process faster and cost-effective
Development Stage Impacted: Antibody Generation and Screening
Solution Impact and Time Savings: EVQLV’s use of AI to accelerate the screening and optimization of antibody candidates directly impacts the antibody generation and screening phase.
As mentioned, traditional screening processes can extend up to 18 months due to the labor-intensive nature of evaluating millions of potential candidates.
The long, nonlinear process is expensive, taking over a decade and billions of dollars per approved antibody. Despite substantial investment, success rates are low due to the complexity of human immune systems.
However, the EVQLV platform streamlines antibody design using AI optimization. Rather than guessing sequences and discovering viable antibodies, EVQLV computationally designs optimized candidates.
It can rapidly predict and enhance properties like target binding affinity by modeling sequence-function relationships. This customizable in silico design minimizes the guesswork, assays, and failure-prone cycles needed.
EVQLV has continuously demonstrated the ability to select promising antibody leads faster and more efficiently than traditional discovery. Replacing lengthy iterations with AI-powered design can accelerate antibody development and boost success rates.
Its CEO, Andrew Satz, said they have “progressed from analyzing 10,000 antibodies in six months to a million in 24 hours and aiming for a billion”.
Also Read: Top 10 Startups Leading in Drug Discovery Using AI
Future Outlook
While still in its early stages, AI in antibody development has the immense potential to revolutionize the field and offer significant benefits for researchers and patients.
For researchers, AI promises faster development timelines, potentially reducing the process from months to days, and broader exploration capabilities, allowing them to analyze 100 times more antibody candidates than traditional methods. This translates into the potential for faster breakthroughs and exploration of previously “undruggable” targets.
Further, AI-driven development can lead to 20% lower costs for antibody therapies, making them more accessible to patients globally.
Additionally, AI’s ability to tackle previously deemed “undruggable” targets holds the potential to unlock treatments for 50% more diseases currently lacking effective options.
These potential benefits highlight the transformative power of AI in accelerating discovery, expanding treatment possibilities, and ultimately improving healthcare outcomes for countless individuals.
As per Andrew Satz, Founder of EVQLV, “AI’s strength lies in its ability to drive down costs, much like how a smartphone consolidates numerous devices from the past into one. It’s an optimistic view, but I believe in the possibility of starting a biotech company with less than $100,000. That would indeed be revolutionary.”
As AI-driven methodologies gain traction and regulatory bodies adapt, novel therapeutics will become significantly more available, offering new hope to patients worldwide.
Contact us for an in-depth view of startups using AI in antibody development, their market potential, and their technology readiness level:
Author: Simran Gohan and Vikas Jha, Solutions.
