When Graph AI Meets Generative AI: A New Era in Scientific Discovery

In recent years, artificial intelligence (AI) has emerged as an important tool in scientific discovery, opening new avenues for research and accelerating the pace of innovation. Of the various AI technologies, Graph AI and Genative AI are especially useful because of their potential to transform the way scientists approach complex problems. Individually, each of these technologies has already made significant contributions in diverse areas such as drug discovery, materials science and genomics. But when combined, they create an even more powerful tool for solving some of science’s most challenging questions. This article explores how these technologies work and combine to drive scientific discovery.

What are Graph AI and Generative AI?

Let’s start by breaking down these two technologies.

Graph AI: the power of connections

Chart AI works with data represented as networks or graphs. Think of nodes as entities, such as molecules or proteins, and edges as the relationships between them, such as interactions or similarities. Graph Neural Networks (GNNs) are a subset of AI models that excel at understanding these complex relationships. This makes it possible to discover patterns and gain deep insights.

Graph AI is already used in:

• Drug discovery: Modeling molecule interactions to predict therapeutic potential.
• Protein folding: Decoding the complex shapes of proteins is a long-standing challenge.
• Genomics: Mapping how genes and proteins relate to diseases to discover genetic insights.

Generative AI: creative problem solving

Generative AI models, such as large language models (LLMs) or diffusion models, can create entirely new data, including text, images, or even chemical compounds. They learn patterns from existing data and use that knowledge to generate new solutions.

Major applications include:

• Designing new molecules for drugs that researchers might not have thought of.
• Simulating biological systems to better understand diseases or ecosystems.
• Propose new hypotheses based on existing research.

Why combine the two?

Graph AI is good at understanding connections, while Generative AI focuses on generating new ideas. Together they provide powerful tools to tackle scientific challenges more effectively. Here are some examples of their combined impact.

1. Accelerating drug discovery

Developing new drugs can take years and cost billions of dollars. Traditionally, researchers test countless molecules to find the right one, which is both time-consuming and expensive. Graph AI helps by modeling the interactions of molecules, narrowing down potential candidates based on how they compare to existing drugs.

Generative AI fuels this process by creating entirely new molecules designed for specific needs, such as binding to a target protein or minimizing side effects. Graph AI can then analyze these new molecules and predict how effective and safe they might be.

For example, in 2020, researchers used these technologies together to identify a drug candidate for the treatment of fibrosis. The process took just 46 days – a vast improvement over the years it normally takes.

2. Solving protein folding

Proteins are the building blocks of life, but understanding how they fold and interact remains one of the most difficult scientific challenges. Graph AI can model proteins as graphs, mapping atoms as nodes and bonds as edges, to analyze how they fold and interact.

Generative AI can build on this by suggesting new protein structures that may have useful properties, such as the ability to treat disease. A breakthrough came with DeepMind AlphaFold used this approach to solve many protein folding problems. Now the combination of Graph AI and Genative AI is helping researchers design proteins for targeted therapies.

3. Promotion of materials science

Materials science searches for new materials with specific properties, such as stronger metals or better batteries. Graph AI helps model the interactions between atoms in a material and predicts how small changes can improve its properties.

Generative AI goes one step further by proposing completely new materials. These may have unique properties, such as better heat resistance or improved energy efficiency. Together these technologies are scientists help creating materials for next-generation technologies such as efficient solar panels and high-capacity batteries.

4. Uncovering genomic insights

In genomics, understanding how genes, proteins and diseases are linked is a major challenge. Graph AI maps these complex networks, allowing researchers to discover relationships and identify targets for therapy.

Generative AI can then propose new genetic sequences or ways to modify genes to treat diseases. For example, it can suggest RNA sequences for gene therapies or predict how genetic changes might affect a disease. Combining these tools will accelerate discoveries and move us closer to treatments for complex diseases such as cancer and genetic disorders.

5. Knowledge discovery from scientific research

A recent study of Markus J. Buehler shows how a combination of Graph AI and Genative AI can discover knowledge from scientific research. They used these methods to analyze more than 1,000 articles on biological materials. By building a knowledge graph of concepts such as material properties and relationships, they discovered surprising connections. For example, they found structural similarities between Beethoven’s 9th Symphony and certain biological materials.

This combination then helps them create a new material: a mycelium-based composite, modeled after Kandinsky’s artwork. This material combined strength, porosity and chemical functionality, demonstrating how AI can drive innovations across disciplines.

Challenges and what’s next

Despite their potential, Graph AI and Genative AI have challenges. Both require high-quality data, which can be difficult to find in fields such as genomics. Training these models also requires a lot of computing power. However, as AI tools improve and data becomes more accessible, these technologies will only get better. We can expect them to make breakthroughs in many scientific disciplines.

The bottom line

The combination of Graph AI and Genative AI is already changing the way scientists approach their work. From accelerating drug discovery to designing new materials and unraveling the mysteries of genomics, these technologies are enabling faster, more creative solutions to some of science’s most pressing challenges. As AI continues to evolve, we can expect even more breakthroughs, making it an exciting time for researchers and innovators alike. The merger of these two AI technologies is just the beginning of a new era in scientific discovery.