In this interview, News Medical speaks with Mohammed Akhlaq at LabGenius Therapeutics about how automation and machine learning are pioneering the discovery of complex antibodies following the company’s recent expansion of its EVA™ platform.
To start, can you tell us a little about your role at LabGenius Therapeutics and what excites you most about working on automation in antibody discovery?
At LabGenius Therapeutics, I’m focused on applying my expertise and experience in automation and assay development to increase our experimental throughput while maintaining data quality.
This is exciting because this high-quality, disease-relevant data fuels our active learning process. The more experimental data we generate, the stronger our ML-enabled antibody discovery process becomes.
LabGenius has recently expanded its platform with a high-throughput workcell. Can you walk us through what this expansion involved and what makes this new system unique?
I’ll start by giving a brief overview of the platform. The EVA™ platform works by combining active learning with automated functional screening in a closed loop. This approach, which is largely free from human bias, allows us to explore large areas of antibody design space and ultimately discover high-performing molecules, often with non-intuitive designs.
The recent expansion further automates the wet-lab steps, enabling us to produce and purify greater numbers of multispecific/multivalent antibodies and assess their characteristics in the final format using disease-relevant, cell-based assays.
With our platform, we’re now able to design, produce (in mammalian cells), purify, and characterize panels of up to 2,300 multispecific/multivalent antibodies in just 6 weeks. We believe this capability to be world-leading.
Scientist explains how AI can accelerate medical discovery
You’ve tripled experimental throughput with this expansion. What specific technical improvements or automation advancements made that possible?
Working with Beckman Coulter as our preferred integration partner has been instrumental — bringing together a considered selection of over 33 devices onto a single platform is no trivial task.
The partnership has enabled us to build a highly complex molecular biology workflow that delivers purified and sequence-verified DNA, ready for mammalian cell transfection.
Primarily, the utilization of large labware and reagent storage capacity, coupled with state-of-the-art automated liquid handling technologies — such as the Echo acoustic dispensing and Bomek i7 liquid handling robot — gives us the ability to carry out near 24/7 process run time.
Increasing experimental throughput directly strengthens your machine learning-driven discovery process. Can you explain how this additional data improves the predictive power of your models?
The more antibody designs we can evaluate experimentally in the wet lab, the more ML-grade data we can feed our models. This not only increases the accuracy of our models but also expands the size of the design space that we can explore in silico. This is important because many high-performing molecules have non-intuitive designs.
In other words, there often isn’t an obvious relationship between a molecular design and its function, so many of our high-performing molecules wouldn’t have been discovered by a human protein engineer.
Scaling up data generation is one part of the equation, but how do you ensure that the data remains high-quality and disease-relevant as throughput increases?
As throughput increases, it’s important that we maintain the integrity of the data we collect.
We achieve this by automating not only the hundreds of discrete experimental steps in the wet lab but also the computational steps that bookend the process. For instance, all experimental data are automatically uploaded to the cloud and processed using purpose-built data pipelines that address processes like QC, normalization, and curve fitting.
This reduces the chances of introducing human error and frees up our scientists’ time to focus on more fruitful experimental work.
With this increased capacity, LabGenius can now run more antibody discovery programs in parallel. How does this impact the speed and efficiency of bringing potential therapeutics to development?
Although the EVA™ platform is modality- and format-agnostic, our internal pipeline focuses on developing selectivity-enhanced antibodies for the treatment of solid tumors. With the expansion of the new platform, we are able to run multiple lead optimization programs simultaneously. This means we can collaborate with pharmaceutical partners on joint programs while also keeping the momentum behind our wholly-owned pipeline of assets.
What were some of the biggest challenges in scaling up the platform, and how did you and your team overcome them?
Putting together such a sophisticated platform inevitably has its challenges. As we anticipated, the biggest hurdle was ensuring its mechanical and biological robustness. Working collaboratively with the team at Beckman Coulter enabled rigorous testing to ensure reproducible results.
Another challenge was automating the colony-picking process using the integrated Amplius imaging system and the Biomek i7 robot.
The ability to pick colonies printed by the Echo accurately and precisely has been overcome with the help of excellent scripting support by the Beckman applications team. This enables efficient colony picking and inoculation, which enables the completion of the molecular biology process within 7 days for 2,300 designs.
Image Credit: LabGenius Therapeutics
Looking ahead, how do you see automation and machine learning continuing to shape the future of antibody discovery?
As antibodies become more complex, the relationship between their design and function will become less intuitive.
Our high-throughput characterization data, when combined with ML, allows us to co-optimize these increasingly complex multispecific/multivalent antibody formats with sophisticated modes of action, ultimately setting a new standard for safety and efficacy.
Are there any upcoming innovations or improvements you’re particularly excited about in LabGenius’ pipeline?
We’re really excited by our molecules’ ability to overcome on-target, off-tumour toxicity, a drawback too often associated with approved anti-tumour therapies. By leveraging avidity-driven selectivity, the antibodies in our pipeline can differentiate between healthy and diseased cells based on differential TAA expression.
This is exciting because, with conventional methods, we see only a marginal widening of the therapeutic window, whereas our approach achieves complete on/off killing selectivity. (Watch our recent data presentation at PEGS here).
As LabGenius continues to scale its capabilities, what do you see as the next big milestone for the EVA platform?
There are always new opportunities to enhance our platform and keep up with the latest scientific innovations. The next milestone we are working towards is the onboarding of increasingly more sophisticated antibody designs — for example, we’re currently onboarding tri- and tetra-specific antibodies with complex modes of action.
Learn more about LabGenius here!
Where can readers find more information?
About Mohammed Akhlaq
As Head of Automation, Mohammed leads a team of automation scientists who develop precision-driven robotic and laboratory automation processes.
Mohammed joined LabGenius from Charles River, where he specialized in the development and automation of complex assays. Prior to that, he spent 23 years at Novartis, refining high-throughput screening methods.
Researchers uncover cause of persistent blood clots despite treatment