One was pizzazz. Although Chemaxon is a well-established, professional company, it certainly is not staid: it never fails to give the image of youth, confidence, and vitality in its approach to meeting production. The lighting was vivid, with shades of pink and purple predominating this time, and speakers were welcomed to the stage with fanfares signaling the appearance of celebrities. There were cameras everywhere: it was like being in a TV studio for a day.

The take-home memory for us, the speakers and panel chair (me), was our mugshots. Not only were the headshots much advertised on the screen and social media, but also we were given framed prints to take away. This is certainly my favorite memento (thanks, Chemaxon, for any photoshopping) and I proudly present it here. The framed version now has pride of place on the executive desk in the Warr office.

Armed with his Catch box “roving microphone”, Gabor Engel of Chemaxon ably chaired the proceedings. This gizmo was new to me; the photos say it all. On a serious note, the slick audiovisual technology and timing equipment also aided Gabor in keeping the whole schedule moving smoothly and to the schedule planned.

But this meeting was not just a miracle of fun, good planning, slickness, and high technology. I really do have to describe briefly the scientific content. The imminent acquisition of Chemaxon by Certara was inevitably at the back of people’s minds, so the welcome and introduction by Chemaxon’s CEO Richard Jones was significant, and some of his comments were taken up later by Adrian Stevens (Chief Product Officer).

Richard Jones and Gabor Engel

Richard remembered how 25 years previously, Ferenc Csizmadia the founder of the company had talked about the accessibility of Marvin. By 2022, Chemaxon has come to be recognized as the best in drawing, search, compliance, collaborative research and design, calculators, standardization, and registration.

There are about 500,000 users of Chemaxon software in academia and education and 750,000 users overall in about 600 active organizations in more than 130 countries. The company employs 240 people, in five offices, and has six distributors in Asia-Pacific. The goal is to be the number one cheminformatics company by maintaining and expanding the best-in-class chemistry building blocks and combining them to help you create the best chemistry cloud platform.

And this is not going to change! With Certara AI, the company can combine these with D360 and Simcyp to provide the solution that can analyze data across multiple areas and create more chemistry-intelligent ideas for novel targets and give you a higher chance of success in predicting clinical trial outcomes.

Timur Madzhidov

Timur Madzhidov of Elsevier presented work that his company did in collaboration with Janssen. They built an ML-optimized Reaxys flat file as an extension of the Reaxys Flat File (RFF) to support AI and ML initiatives. Standardization for ML applications might be different from that used for data representation. Data representation differs in databases, “chemical beauty”, and cheminformatics software capabilities. Using the RFF, the team were able to reduce the time spent on data preparation by 80-90%. Fifteen percent more reactions were ready to use with ML/AI models and the models were improved due to a larger modeling dataset and greater homogeneity of data.

Josef Eiblmaier

The next speaker was Josef Eiblmaier of Pharmalex (a Cencora company). Data collection and preparation accounts for 80% of the effort of a data scientist; 20% is mining etc., according to Forbes. Josef reported on successful work carried out in collaboration with Evotec to build a centralized data access platform. The One Data Access vision was to design and implement ONE centralized architecture and ONE solution for global scientific data access and predictive modeling.

Karl-Heinz Baringhaus

Karl-Heinz Baringhaus of Sanofi (pictured at the end of the day) was forward-looking and holistic in his presentation, which considered the whole pharma value chain. His solutions for grasping the opportunities for better drug discovery included:

• understanding disease (then target identification)
• new technologies, of which AI/ML is only one
• AI in drug discovery
• self-driving labs
• explainable AI (building trust)
• better decision making
• using advanced data analytics
• incorporating model-informed drug discovery and development
• mitigating bias
• collaborative research and open data sharing among pharmaceutical companies, academic institutions, and regulatory bodies, and
• continuous learning and adaptation.

Becky Upton

Becky Upton showed some results from the Pistoia Alliance’s Lab of the Future survey, suggesting that the three biggest barriers to making the best use of experimental data are unstructured data, data silos or no access to data, and lack of metadata standardization. The biggest barriers to implementing AI/ML at scale are low quality, poorly curated datasets, data that are not FAIR, and privacy and security concerns around sensitive data. Becky discussed the role of ontologies and described the Pistoia Alliance’s current and upcoming ontologies.

Jessica Lanini

Jessica Lanini of Novartis has published a perspective¹ with colleagues from Novartis, Saarland University, ETH Zurich, AstraZeneca, and Eindhoven University of Technology. Academia and industry are both driving the field of molecular ML research. Models have permeated almost every step in the DMTA cycle at Novartis. Due to the fast emergence of new ML algorithms, the field needs to adapt quickly, including changes in collaboration (sharing data, protocols, code, and models) and multidisciplinary scientists’ education. More collaboration efforts between academia and industry to share data or code might lessen the gap between exploratory and applied research work. There are private-public collaborations in which science can be advanced in real-world project set-ups while keeping sensitive data private.

Thrasyvoulos Karydis

Thrasyvoulos Karydis discussed Deepcure’s interactive platform which allows chemists and AI to co-create while avoiding human bias. AI can innovate only if you expand the relevant chemical space and remove human and data bias towards compounds that have already been made. With its MolGen and PocketExpander solutions, Deepcure addresses this challenge.

Peter Ertl

Peter Ertl, recently retired from Novartis, is well known for his work on scaffolds^2,3 His talk concerned his “magic rings”. You can try out his software for free at https://ertlmolecular.com.

Adrian Stevens

Adrian Stevens of Chemaxon is a good “storyteller” and he presented concrete examples for his ideas on how to help future drug discovery needs. A particularly memorable quip was “You call it AI when you want your CEO to fund your project. You call it ML when you talk to colleagues. And if you’re old like me, you call it QSAR”. Some of Adrian’s topics were pulling data from D360 into Design Hub; supporting complex modalities; sketching, rendering and publishing chemical drawings with Marvin (calculating properties while you sketch); and handling chemistry in biologics.

Nessa Carson

Nessa Carson of AstraZeneca aims to save chemists from wasting their time. Digitalization is one of the most prominent issues for chemical industry, according to CEOs. Nessa discussed data and the digital ecosystem. Data flow enables enabling technology. Data connection is starting to happen in self-driving labs, reaction prediction, smart instruments, and DMTA acceleration. Everybody needs to think about the user interface. Users worry about change, but user-friendliness leads to adoption. Nessa’s motto was “Your product doesn’t work if the user can’t work it”.

Karl-Heinz Baringhaus, Becky Upton, Adrian Stevens, Jeremy Frey and me

Now I will digress and summarize the panel discussion which took place at the end of the morning session. Karl-Heinz Baringhaus, Becky Upton, Adrian Stevens, and Jeremy Frey discussed the future of R&D and the trends, challenges, and solutions transforming the way we work. We did not confine the panelists to technological solutions alone.

Since I chaired the discussion, I was unable to make notes on the 45 minutes of full and frank discussion. I did ask the panelists in advance for some keywords or phrases that represented the issues they were likely to address. They suggested topics such as artificial intelligence in all areas of drug discovery and development, data access, standardization, quality, efficient collaborations, automation and self-driving labs, the role of people, human-centric research, and many more. These are all complex topics, so in the end, we managed to touch on only a handful of them in our discussion. Much of our discourse centered on the importance of data quality, the trust in models, trust in data, and as a result the applicability of AI in research and its future role in decision making.

Jeremy Frey

It was good that Jeremy Frey of Southampton University was the last speaker of the day because his presentation was truly visionary and gave us much to go away and discuss. There is no way that I can fairly cover his talk in just a few sentences. He talked about the progress of digitalization of his research in the eras defined by “pre-digital computers”, “computers are useful”, “computers are essential” and “computers (perhaps) take over”. Much of the talk did cover augmented intelligence (but not “artificial” intelligence). One of his conclusions was that augmented intelligence (AI) needs information architecture (IA).

Jeremy quoted from Peter Day’s “The Philosopher’s Tree: a Selection of Michael Faraday’s Writings”:

“Faraday’s laboratory notebooks are also remarkable in the amount of detail that they give about the design and setting up of experiments, interspersed with comments about their outcome and thoughts of a more philosophical kind. All are couched in plain language, with many vivid phrases of delightful spontaneity….”

I like to think that Jeremy’s own notebooks, projects, and presentations have much in common with Faraday’s.

The meeting concluded with more networking opportunities at a pleasant reception. Since I was involved in the organization of this meeting, obviously I am really happy that it turned out so successful, as evidenced by the number of attendees and the positive nature of their feedback.

The Chemaxon team

Most importantly, I have to conclude by thanking and praising all the people who really did the hard work: the vibrant and enthusiastic members of the Chemaxon team. I look forward to the “Certainty” of a happy and successful Chemaxon-Certara unification.

(1)        Volkamer, A.; Riniker, S.; Nittinger, E.; Lanini, J.; Grisoni, F.; Evertsson, E.; Rodriguez-Perez, R.; Schneider, N. Machine learning for small molecule drug discovery in academia and industry. Artif. Intell. Life Sci. 2023, 3, 100056.

(2)        Ertl, P. Magic rings: navigation in the ring chemical space guided by the bioactive rings. J. Chem. Inf. Model. 2022, 62 (9), 2164-2170.

(3)        Ertl, P. Database of 4 million medicinal chemistry-relevant ring systems. J. Chem. Inf. Model. 2024, 64 (4), 1245-1250.