Five years ago, Google DeepMind’s AlphaFold fundamentally changed how we understand biology. By solving the 50-year-old protein folding problem, it opened doors that scientists had been pushing against for decades. Today, as we stand on the shoulders of this breakthrough, a new question emerges: What comes after prediction?

The answer may lie not in better predictions of individual components, but in understanding how those components work together as integrated living systems. This is the premise behind GenBio AI’s AI-Driven Digital Organism (AIDO), which represents a conceptual shift from molecular prediction to systems-level biological intelligence.

AlphaFold’s Revolutionary Impact

AlphaFold 2’s impact on structural biology has been profound. With structural predictions for over 240 million proteins, it has empowered more than 3.3 million scientists across 190 countries. The achievement was recognised with the 2024 Nobel Prize in Chemistry, underscoring that a computational method had fundamentally transformed experimental biology.

But AlphaFold was a starting point, not the finish line. Google DeepMind has since developed an ecosystem of specialised AI tools, each addressing different aspects of molecular biology.

The DeepMind Biological AI Ecosystem

Why These Tools Matter

Each of these models represents a significant advance in applying AI to specific biological challenges. Collectively, they provide accuracy that supports experimental design, the ability to analyse millions of molecules in hours rather than decades, and accessible tools that democratise advanced biological research. These are not incremental improvements. They are fundamental capabilities that did not previously exist, now available to researchers worldwide.

The Limits of Individual Specialization

Here is the fundamental challenge: biology does not operate in isolated modules.

In a living cell, DNA transcription, RNA processing, protein folding, cellular expression, and evolutionary pressures are deeply interconnected parts of a complex, dynamic system where everything influences everything else.

Consider drug development as an example. You might use AlphaFold 3 to predict how a drug candidate binds to its target protein with high accuracy. But the next set of questions is harder: How does that binding affect the protein’s function in its cellular context? What downstream effects cascade through regulatory networks? How do genetic variants in different populations affect efficacy? What about off-target effects in other tissues and cell types?

Current specialised models excel at the first question. The remaining questions require understanding the system as a whole. This is not a criticism of specialised models; it is an acknowledgment that we need complementary approaches to address different levels of biological organisation.

AIDO: Integrated Systems Biology

GenBio AI’s AI-Driven Digital Organism (AIDO) takes a different approach. Rather than optimising for specific prediction tasks, AIDO is designed to mirror biology’s inherent interconnectedness.

The Architecture of Integration

AIDO comprises six foundation models working in concert: a DNA model with 7 billion parameters trained across 796 species; an RNA model capturing transcriptional dynamics; a protein sequence model for functional prediction; a protein structure model for three-dimensional folding; a single-cell expression model for understanding cellular state; and an evolutionary information model capturing conservation, selection, and phylogenetic relationships.

What makes AIDO distinctive is not the existence of multiple models, but how they communicate through hierarchical representation propagation and continual pretraining. Information flows between scales, allowing molecular-level understanding to inform cellular predictions and vice versa.

Unified Task Handling

AIDO can address up to 300 diverse biological tasks simultaneously. Traditional AI models are typically optimised for one or two specific tasks and perform well on those, but they struggle when reasoning across scales or integrating different types of biological information. AIDO’s architecture is specifically designed for this cross-scale reasoning.

The stated long-term goal is to move beyond prediction toward biological programming: not just understanding “what is” but designing “what could be.” This requires understanding the design principles governing how components assemble into functional systems and how robustness emerges.

Why Both Approaches Are Essential

The future of biological AI is not about choosing between specialised models and integrated systems. Both are necessary and complementary.

Specialised models give us the vocabulary of biology: accurate descriptions of individual elements. Systems models aim to provide the grammar: how those elements combine to create functional meaning.

Real-World Implications

The convergence of these approaches opens concrete possibilities across multiple domains.

Personalized Medicine

Consider predicting not only whether a patient carries a pathogenic genetic variant, but how that variant propagates effects through cellular networks, how it interacts with other variants in their genome, and how treatments will perform in their specific biological context. This requires both accurate molecular predictions and systems-level integration.

Drug Discovery

Beyond predicting drug–target binding, we could simulate a drug’s effects throughout relevant cell types, predict off-target effects, understand resistance mechanisms before they emerge, and design combination therapies that account for systems-level responses.

Synthetic Biology

Designing organisms or cellular systems with novel capabilities requires understanding not just individual components but how to assemble them into functional, robust systems. We need both the parts list from specialised models and the assembly logic from systems models.

Disease Understanding

Many diseases arise not from single molecular defects but from dysregulation of biological networks. Understanding and treating these conditions requires systems-level thinking informed by accurate molecular predictions.

The Path Forward

We are at an inflection point in computational biology. Specialised models have given us unprecedented ability to characterise biological components. Systems-level approaches like AIDO are beginning to explore how those components give rise to emergent properties of living systems.

The next breakthroughs will likely come from the interplay between these approaches: using specialised models to generate high-quality training data for systems models; using systems models to identify which molecular predictions matter most; developing hybrid architectures that combine task-specific accuracy with cross-scale reasoning; and creating feedback loops where experimental validation improves both types of models.

This is not just about building better computational tools. It is about developing a quantitative, predictive framework for understanding life at every level of biological organisation. Neither specialised nor systems-level approaches alone are sufficient. Together, they are moving us toward a digital representation of biology that can predict, simulate, and ultimately enable us to engineer living systems.

We are still in the early chapters of this story. The question is not which approach will prevail, but how quickly we can learn to use them in concert to address the biological challenges that matter most.

Research funding in Europe is currently more in the spotlight than ever before. With the Horizon Europe programme, the European Union is investing billions of euros in innovation, sustainability, healthcare, and technology. As the 2021–2027 funding period nears its end, it is time to analyse the global distribution of funds, cooperation among institutes, and research topics to support well-informed decisions for both scientists and the EU Commission as they plan for the next funding period.

The analysis of the Horizon Europe datasets provided valuable insights. The datasets contain information on countries, institutes, their funded projects, and maximum awarded amounts. The core findings, enhanced with interactive visualisations and rankings, are summarised below.

Which Countries Participate in Horizon Europe Funded Projects?

Very few countries worldwide have no involvement in European Horizon-funded projects. Russia is one of them. All other countries are colour-coded in the interactive map below. The colour scale is based on the sum of funded projects per country, normalised by the number of funded institutes per country.

The highest project involvement per institute is found in European countries such as Greece and Denmark. There are, however, visible differences in project involvement between Eastern and Western Europe. Horizon Europe funded projects also include significant participation from countries across North America, Australia, and Africa, highlighting the programme’s global reach and international collaboration.

Which Research Institutions Are Most Frequently Involved?

The bar chart below shows the absolute number of project participations per institution. Each bar is colour-coded by country.

Funding is concentrated among a few major players that hold leading positions across Europe. The French National Centre for Scientific Research (CNRS) leads by a wide margin, followed by the Spanish National Research Council (CSIC) and Germany’s Fraunhofer Society in second and third place. Notably, Denmark, a comparatively small country, ranks highly with three institutes among the top 20.

What Are the Funded Research Projects About?

To answer this question, keywords submitted with the proposals were assessed. The word cloud below shows the top 100 keywords, with the size of each word corresponding to its frequency of occurrence.

AI research projects are heavily funded, reflecting the considerable attention this topic received during the most recent proposal evaluations. It is encouraging that sustainability and circular economy also rank highly. However, it is surprising to find so few keywords related to life sciences.

Which Institutions Receive the Highest Funding?

The colour coding in the chart below reflects the countries of the institutes.

The CNRS leads by a wide margin, securing approximately €2.55 billion in total funding. The major Spanish and Italian research institutions occupy the next top ranks.

The Catholic University of Leuven (KU Leuven) stands out as a top recipient of EU research funding, reflecting its exceptional ability to secure substantial grants under Horizon Europe. With more than €170 million awarded across 284 projects, as reported by the university in 2023, KU Leuven owes its success to a well-developed institutional support system, including dedicated grant writing assistance and strong policies fostering innovation. Closely following is the Technical University of Denmark (DTU), which similarly benefits from robust participation in Horizon Europe and other collaborative research initiatives. Both institutions demonstrate how strategic support structures and focused research expertise translate into competitive funding success.

From Germany, the Fraunhofer Society and the Max Planck Society also rank high, reflecting their prominent role in European research. Following these are smaller universities such as Aarhus (Denmark), Ghent (Belgium), and Lund (Sweden), which are notable for their strong research output relative to their size.

Which Countries Receive the Highest Funding?

A review of the funding allocation by country clearly indicates that Western European nations, specifically Spain, France, Italy, and Germany, receive the highest levels of financial support. This is largely attributable to the high concentration of research institutes in these countries. Spain, for example, receives up to €50 billion over the entire funding period.

Collaboration Networks

Collaboration is a key driver of research across Europe and the world, and it ensures that funding is allocated effectively. To better understand the research landscape within Horizon Europe-funded projects, a network-based approach was used to highlight crucial partnerships and identify the most collaborative institutions.

The first network below includes the top 50 institutions that appear most frequently in the dataset. These institutions participate across many projects, indicating broad activity in the research ecosystem.

To take a deeper look at the strongest cooperative relationships, the second network was filtered according to the most frequent collaborating institution pairs.

Although many French institutions are centrally located within the network, the networks themselves are highly unclustered. This suggests that European research collaborations extend well beyond the borders of any single country. The leading institutions in France, Spain, Italy, and Germany play a crucial role in maintaining a robust framework for collaborations across Europe. These organisations serve as hubs, enabling and supporting other institutions within the European research ecosystem. Their central position helps ensure strong connectivity and sustained cooperation among diverse partners across the continent.

Conclusion

The analysis makes it clear that Horizon Europe is more than a funding programme; it is a catalyst for networking and interdisciplinary collaboration that extends globally. While large institutes and Western European countries hold a dominant position, specific partnerships are fostering the emergence of new research hotspots.

As the new funding period approaches, the European Commission faces the task of evaluating project topics, while researchers must strategically reconsider their collaborative partnerships.