12 April 2022 in Molecular Pharmaceutics
15 February 2023 in Current Opinion in Structural Biology
28 June 2024 in Cell Biology and Toxicology
Structural cardiotoxicity (SCT) presents a high-impact risk that is poorly tolerated in drug discovery unless significant benefit is anticipated. Therefore, we aimed to improve the mechanistic understanding of SCT. First, we combined machine learning methods with a modified calcium transient assay in human-induced pluripotent stem cell-derived cardiomyocytes to identify nine parameters that could predict SCT. Next, we applied transcriptomic profiling to human cardiac microtissues exposed to structural and non-structural cardiotoxins. Fifty-two genes expressed across the three main cell types in the heart (cardiomyocytes, endothelial cells, and fibroblasts) were prioritised in differential expression and network clustering analyses and could be linked to known mechanisms of SCT. This transcriptomic fingerprint may prove useful for generating strategies to mitigate SCT risk in early drug discovery. Graphical Abstract
28 May 2026 in Frontiers in Immunology
The unwanted immune response to Biologic Therapies can result in anti-drug antibodies that complicate clinical development and may adversely affect patient outcomes. At present, prediction of the impact of this immunogenicity before starting clinical trials is impossible, due to the complexity of the immune system and the multiple factors that contribute to the risk of immunogenicity. Advances in computational methods and power will enable improvements in prediction of immunogenicity. A workshop at EMBL-EBI brought together industry experts and academics to reflect on the contributions of artificial intelligence (AI) and machine learning (ML) to immunogenicity prediction, to review current practices across industry, and to look to future opportunities for applying AI technologies. This review was inspired by the topics and discussions presented at the workshop. Machine learning has been employed for immunogenicity prediction for more than 20 years. Specifically, the prediction of peptides bound by the Major Histocompatibility Complex (MHC) Class II molecule has helped to identify potential T cell epitopes, which can be used for selecting candidates with low immunogenicity risk, informing protein engineering for reducing risk, or informing risk assessments and immunomonitoring during clinical trials. Application of ML algorithms in data rich disease areas such as haemophilia is informative for clinical decision making. ML and other AI techniques require large data sets which have been acquired through consistent methods. A challenge for immunogenicity prediction is the harmonization of preclinical risk assessment assays and the clinical measurements of anti-drug antibodies. With imperfect data, quantitative systems pharmacology (QSP) modelling has been applied to link together the immune system with observations of risk factors, with simulations of clinical trials providing a perspective on the immunogenicity risk. Industry workflows are aligned on application of tools and recognise gaps that need to be filled with additional data and assays. Further innovation in modalities requires extension of the risk assessment paradigm and will demand further innovation in immunogenicity prediction approaches. Finally, we address opportunities for AI/ML to solve key questions and reflect on the challenges in validating the predictive capabilities of new models.
15 October 2026 in Nucleic Acids Research
VHHs, the variable regions of camelid heavy-chain antibodies, are increasingly used as therapeutic, diagnostic, and research tools because of their small size and stability. Despite the growing number of publicly available VHH repertoire sequencing studies, there is no large-scale, web-accessible database of VHH sequences. Here, we present OAS-nano, a database containing over 73 million unique VHH sequences collected from 16 public repertoire sequencing datasets and 1 previously unpublished dataset, spanning alpaca, llama, and Bactrian camel repertoires. Sequence reads were processed using a standardized pipeline to extract high-quality VHH sequences. Experimental metadata and sequence annotations enable users to identify sequences appropriate for their use case. OAS-nano is freely available at https://oas-nano.opig.stats.ox.ac.uk/.