Dr Rietveld and a worldwide interdisciplinary team of collaborators established the BEST-CSP network, an EU-funded initiative that aims to transform how we understand and predict the crystal structures of organic molecules. At the heart of this work lies a deceptively simple question: when a molecule forms a crystal, what exact structure will it adopt, and how stable will that form be under real-world conditions?

This question is not purely academic. Crystalline forms matter enormously in pharmaceuticals, agrochemicals, dyes, and advanced materials. For instance, a drug molecule can form several so-called ‘polymorphs’, which are different crystalline arrangements of the same molecules. Those polymorphs can differ in solubility, stability, toxicity, and even therapeutic effectiveness. Some polymorphs may degrade quickly, while other forms may unexpectedly emerge after a drug is on the market, causing costly recalls. For industry and society alike, being able to predict and control which polymorph is present in a product is vital.

Dr Rietveld and the network are writing a perspective paper outlining a key obstacle: while computational crystal structure prediction – or ‘CSP’ for short – has advanced dramatically in recent years, the lack of reliable benchmark data prevents models from being properly tested. A small and limited dataset known as ‘X23’ had long served this role, but it was inadequate for capturing the complexity of real-world molecules.

To move forward, the community needed a new benchmark: a robust collection of high-quality experimental data, built through the coordinated efforts of many laboratories. This is the mission of the BEST-CSP project, which has now gone beyond vision into practice.

A set of benchmark compounds has been chosen – including the molecules carbamazepine, pyrazinamide, phenylpiracetam, sulfamerazine, and benzophenone. Each of these molecules exhibits polymorphism, meaning that they can exist in several different crystal structures. This makes them challenging but highly relevant test cases. Around these compounds, the BEST-CSP network has been designing studies that bring experiment and simulation into dialogue, each testing and refining the other.

Benzophenone, for example, has been a testing ground for two different strands of BEST-CSP research. One group focused on a technique called differential scanning calorimetry (or ‘DSC’), which measures the melting points and energies of phase transitions. Dr. Fulem from Prague and collaborators found that when different laboratories performed DSC on benzophenone, the results could vary significantly depending on how instruments were calibrated and how samples were handled.

By developing a harmonized protocol, proposed by Dr. Eusebio from Coimbra and Dr Piedade from Lisbon, and repeating the measurements, they were able to reduce variability and produce reliable consensus data. This work not only clarified the behaviour of benzophenone’s stable polymorphs, but also demonstrated how careful inter-laboratory studies can sharpen the precision of routine techniques.

Another team studied benzophenone from a structural point of view. To understand how molecules arrange themselves in a crystal, scientists typically use X-ray diffraction, which gives a detailed picture of where each atom sits.

Traditionally, these structures have been interpreted with a very simple model that assumes atoms are perfect little spheres. This works in many cases, but it can miss important details about how electrons are actually distributed around atoms, which in turn affects the accuracy of the structure. More advanced methods now allow researchers to take into account the true, uneven shapes of electron clouds, giving a more realistic picture of the crystal.

The BEST-CSP network involving Dr. Hoser from Warsaw and Dr. Trisovic from Belgrade collected many high-quality benzophenone crystal structures from different labs and compared results across these methods. By doing so, they not only improved our understanding of benzophenone itself but also created a valuable test case that other scientists can use to judge how reliable these different techniques are.

When the team, involving Dr. Price from UCL and Dr Braun from Innsbruck, studied sulfamerazine, they found that one polymorph of the drug (called form II) will reliably switch into another (form I) when heated. Careful experiments confirmed this change and measured how much energy was involved. But when researchers tried to investigate how the drug behaved in different solvents, they discovered an entirely new crystal form, which they named form V.

This new form looked very similar to form I but was arranged just differently enough to be considered a distinct polymorph. The discovery made the picture of sulfamerazine’s behaviour more complicated than expected and showed why predicting polymorphism is so challenging. Even small changes in experimental conditions can produce surprising new results. Computer models were then compared to the experimental findings, helping the researchers see where the models work well and where they still fall short.

Together, these examples show how Dr. Rietveld’s team is building the benchmark dataset envisioned in their perspective paper. By tackling molecules one by one, with multiple laboratories contributing complementary expertise, they are producing not just raw numbers but carefully vetted, uncertainty-quantified data. With robust datasets, computational chemists can finally test whether their models capture reality, and refine them where they fall short.

What is striking about BEST-CSP is how it redefines collaboration. Instead of isolated labs publishing piecemeal results, the network organises coordinated campaigns where DSC, crystal-structure determination, spectroscopy, and modelling are all applied to the same molecule. Each discipline probes a different dimension of the solid-state landscape, and together they form a complete picture.

The long-term vision is transformative. If crystal structure prediction can be made reliable, industries could anticipate and avoid dangerous polymorphic shifts before a product reaches the market. Researchers could design new materials with confidence, knowing that simulations accurately reflect the stability of proposed structures.

By bringing experiment and simulation together, Dr. Rietveld’s BEST-CSP collaboration is advancing crystal structure prediction, laying the groundwork for more reliable medicines, safer materials, and a deeper understanding of the molecular world.