A research team of Nigerian scholars based at the University of Greater Manchester has been honoured at the 2026 International Conference on Cybersecurity, Digital Forensics and AI Applications (ICCSDFAI), held from 25 to 27 June 2026 at Istinye University’s Vadi Campus in Istanbul, Türkiye.
The team took Third Place in the Best Paper Awards for their study, “Evaluating Artificial Intelligence for Predicting and Interpreting Molecular Toxicity from Chemical Structures to Enhance Early Drug Screening.”
The award-winning authors are Daniel Ndifreke Samuel, Professor Celestine Iwendi, Ogunlola John Olalere, Adedeji Edward Adesola, Afeez Oluwaseun Akande and Blessing Alice Alao-Olatunji, all of the Research and Doctoral College, University of Greater Manchester, Bolton.
Teaching machines to spot a dangerous molecule
The winning paper tackles one of the most expensive problems in medicine: drugs that fail late in development because of unexpected toxicity. By the time a compound reaches clinical trials, years of work and enormous sums have already been committed and safety failures at that stage account for a significant share of losses across the pharmaceutical industry.
The team’s approach was to move that safety check to the very beginning of the pipeline, using artificial intelligence to read a molecule’s chemical structure and estimate its risk before a single laboratory experiment is run.
Working with the publicly available Tox21 dataset, specifically its nuclear receptor androgen receptor (NR-AR) endpoint, covering 1,488 compounds, the researchers converted each molecule into a digital “fingerprint” capturing its structural building blocks, then trained a Random Forest classifier to distinguish toxic from non-toxic compounds.
The model achieved 96 per cent overall accuracy and a ROC–AUC of 0.7666, meaning that when presented with a toxic and a non-toxic compound at random, it correctly ranks the toxic one higher roughly 77 per cent of the time. It proved especially reliable at confidently identifying safe compounds, with precision and recall of 0.98 for the non-toxic class.
Crucially, the team went beyond prediction to interpretation. Rather than delivering a verdict from a black box, their system ranks which molecular substructures drove each decision pointing chemists towards the specific fragments associated with toxic behaviour. That transparency, the authors argue, is what makes such tools usable by regulators and medicinal chemists rather than merely impressive on paper.
The researchers were also candid about the model’s limits. Because toxic compounds are rare in the dataset, the system detects them less reliably than it rules out safe ones, which is a well-documented challenge in toxicology data.
A seasoned hand and a rising generation
The paper brings together early-career researchers and an established voice in the field. Professor Celestine Iwendi, a widely published academic in artificial intelligence and cybersecurity and a senior figure at the Research and Doctoral College, University of Greater Manchester, has mentored a growing cohort of Nigerian and international doctoral researchers in the United Kingdom.
The paper has now been published in the conference proceedings and is available in IEEE Xplore, the digital library of the Institute of Electrical and Electronics Engineers, which indexes the research of engineers and computer scientists worldwide. The study can be accessed at DOI: 10.1109/ICCSDFAI70505.2026.11648038.
A point of national pride
Speaking on the achievement, the team described the award as a testament to what Nigerian researchers can build together and said they hoped the recognition would encourage more young Nigerians into artificial intelligence and the health sciences, fields where the country’s own needs are pressing.
The recognition adds to a growing record of Nigerian scholars contributing at the highest levels of international AI research, at the intersection of machine learning and drug safety.
WATCH TOP VIDEOS FROM NIGERIAN TRIBUNE TV