High-performance computing and advanced AI are no longer the exclusive playground of Big Tech and multinational pharma giants. Immunochem, a biotechnology SME based in Wallonia, is proving that smaller ecosystem players can spearhead industry-defining innovation. Over a three-year strategic pivot, the company has successfully integrated CENAERO’s LUCIA supercomputer into its pipeline to automate candidate generation at a fraction of the cost of commercial cloud providers. As Immunochem sends its first entirely in silico-generated candidates for laboratory validation, they are providing a blueprint for the future of European biotech innovation. We met with Philippe Herman from Immunochem for his insights into this rapidly growing research area.

In-silico New Approach Methodologies (NAMs): Merging Biology and AI as an alternative to animal testing.

Could you introduce Immunochem and its main activities?

Immunochem is a company specialising in biotechnology and life sciences. It focuses on antibody generation, using both classical biological approaches and digital tools that combine artificial intelligence with molecular dynamics calculations to optimise and accelerate the discovery of new therapeutic and diagnostic molecules.

One of our main activities is developing the EVOBodies™ platform to transform antibody design. The long-term objective is to reduce, or even replace, the use of animal experimentation by prioritising a fully computational approach.

We have been working to integrate artificial intelligence into our processes for nearly 5 years. In 2025, we made a strategic pivot: the classical approach is like testing thousands of keys on a lock. Our approach is to observe the lock and machine the key that fits it. We have therefore moved from massive screening—producing millions of variants in the hope of identifying an effective one—to targeted design, where AI directly proposes the most relevant candidates.

Unlike traditional immunisation or screening methods, which are often lengthy and uncertain, this approach enables the production of precise, reproducible molecules in a matter of weeks rather than months.

Immunochem’s added value also lies in its ability to experimentally validate antibodies rather than merely design them computationally.

Breaking Down the Scientific and Industrial Hurdles

What are the major scientific or industrial challenges you are currently facing?

One of the major scientific challenges in the field concerns predicting molecular specificity. The challenge is to have computational tools capable of accurately determining whether an antibody will bind to a given target molecule. This is a fundamental scientific challenge.

To address this, we have developed three complementary pipelines: generation, filtering based on biophysical metrics of the interaction, and, before experimental validation, we employ molecular dynamics as an in silico physical validation step to assess the stability of antibody-antigen complexes.

The industrial challenge is equally important: demonstrating that a fully computational pipeline can produce functional molecules that are validated in the laboratory, reproducibly and at a competitive cost.

The ideal objective is to identify the most relevant candidates without having to experimentally test each molecule individually. Moving away from animal immunisation processes also offers several advantages: reducing animal experimentation, saving time, and lowering the costs associated with reagents and experimental campaigns.

The Shift to High-Performance Computing

When did high-performance computing become necessary for your activities?

High-performance computing became essential from the project's outset to overcome the limitations of traditional computing infrastructure and enable the parallelisation of complex models. Some simulations, such as molecular dynamics, require substantial computing capacity. This type of computation cannot be performed efficiently on standard desktop computers and requires, in particular, the use of GPUs and dedicated HPC infrastructure.

In the context of Immunochem’s activities, the HPC approach therefore proved essential.

Overcoming the Limitations of Standard Computing

What types of problems could not be addressed without HPC?

HPC plays a decisive role in accelerating molecular discovery. Without these computing capabilities, some problems involving massive volumes of simulation or modelling would simply take too long to process.

When we target an antigen, we generate and evaluate hundreds of candidate backbones. Each generation involves running a deep learning model, followed by multi-criteria scoring. Without GPU acceleration, a single project would take weeks instead of a few days.

In other words, discovery cycles would be considerably longer, and some approaches based on intensive simulation would become difficult to exploit in practice.

"Relying exclusively on commercial cloud services such as AWS or Google Cloud would be two to three times more expensive than using the LUCIA infrastructure"

Timeline and Economic Gains: The Supercomputing Advantage

What benefits do you observe from using HPC?

As mentioned earlier, the most obvious benefit concerns research and development time.

In a classical animal immunisation approach—for example, in mice—an experimental campaign can take around six months. In the case of nanobodies, the process takes even longer: sometimes it takes three to four months just to obtain the animal, then six to nine months for the immunisation phase, followed by another three months for screening. Altogether, a “classical” project can therefore extend beyond a year.

Using our computational approach on CENAERO’s LUCIA supercomputer, we aim to generate and validate an antibody in approximately three months. The in silico generation phase itself takes only a few days.

The second benefit is economic. For an equivalent computing volume, relying exclusively on commercial cloud services such as AWS or Google Cloud would be two to three times more expensive than using the LUCIA infrastructure. For a biotech SME such as Immunochem, this difference is decisive.

The third benefit is capacity. With a small team, we can work on a large number of targets per year, a volume that would be unthinkable with purely experimental approaches.

The Skills Prerequisite: Bridging Biology and Code

What in-house skills are needed to use HPC effectively?

At a minimum, you need someone proficient in Linux, scripting (e.g., Python and Bash), and job management systems (e.g., SLURM). In our case, our modelling and structural bioinformatics engineer fills this role. He designs the pipelines, manages job submissions on LUCIA, and ensures the connection between computational results and their biological interpretation.

What is important to emphasise is that the learning curve is shorter than people might think. A bioinformatician or data scientist with basic command-line skills can become operational in a few weeks, especially since CENAERO’s technical team provides high-quality support for getting started with the infrastructure.

The real prerequisite is not so much pure HPC expertise as the ability to bridge biology and computation: understanding why a given calculation is being launched, how to interpret the results, and when a computational result deserves laboratory validation.

"HPC enables us to automate and parallelise candidate generation and filtering, thereby considerably increasing our scientific productivity"

Transforming the Biotech Business Model

What is the impact of HPC on Immunochem’s capacity for innovation?

HPC clearly acts as a catalyst. It allows us to automate and parallelise a large number of tasks, thereby increasing our scientific productivity considerably.

With a relatively small team, we can work on a large number of targets per year. In addition, the use of dedicated HPC infrastructure helps keep costs reasonable. For an equivalent computing volume, relying exclusively on commercial cloud services could cost two to three times as much.

In concrete terms, HPC enables us to automate and parallelise candidate generation and filtering, thereby considerably increasing our scientific productivity. Where a traditional laboratory tests a few dozen molecules per campaign, we evaluate hundreds in silico before sending only a handful for experimental validation.

This approach also changes our business model. Instead of investing heavily in reagents and laboratory time for random screening, we focus our resources on the most promising candidates. The cost-to-result ratio is radically improved.

A Call to Action for Biotech SMEs

What message would you like to convey to biotechnology stakeholders?

AI and HPC are no longer reserved for big pharma or the Big Tech companies. The tools are available—many of them open source—and computing infrastructures such as LUCIA make these technologies accessible to Walloon SMEs, provided they have the right expertise.

What is often missing is the first step. There is a natural tendency to regard commercial cloud services as simpler and more accessible. Our experience shows the opposite: CENAERO’s technical support is excellent, costs are controlled, and once the pipeline is in place, the return on investment is immediate.

My advice to biotech stakeholders who are hesitating is this: start with a pilot project. Identify a concrete problem—structure prediction, affinity scoring, sequence analysis—and test it on LUCIA. You will be surprised by how quickly you can become operational.

"Convergence of biotechnology, AI, and HPC is redefining our domain, and Wallonia has the assets to be at the forefront of this transformation"

Looking Ahead: The Biotech Revolution in Wallonia

In three years, Immunochem has moved from a classical physical library approach to a pipeline in which AI designs the molecules, and the laboratory validates them. This pivot would not have been possible without two elements: the Walloon ecosystem, with Biowin, CENAERO, and our academic partners; and the conviction that high-performance computing is a strategic lever, not a technological gimmick.

We are currently sending our first in silico-generated candidates for experimental validation. The coming months will tell us whether our bet was the right one. But one thing is certain: the convergence of biotechnology, AI, and HPC is redefining our domain, and Wallonia has the assets to be at the forefront of this transformation.

Interview by Frederique Jacobs for EuroCC Belgium