These pages contain links to various tools and resources created by the staff and students associated with the two Biomedical Artificial Intelligence Centres for Doctoral Training funded by UKRI at the University of Edinburgh.

Courses and Training

• Introduction to Python for Biomedical Innovation. A student developed crash course in using Python for biomedical data analysis. This course is presented in modules including coding notebooks to allow you to develop practical skills.
• An Introduction to Multi Modal Analysis Using Networks. This workshop was run at the Intelligent Systems for Molecular Biology (ISMB) confrence in Montreal in July 2024. It is a Python based hands on tutorial introducing network methods including network similarity fusion (SNF) and the use of Graph Neural Networks (GNNs) to perofrm classification tasks on patient specific networks (PSNs).
• How to Create a Protein. A short course on proteins and protein design aimed at early career students.

Public Outreach Activities

• Zoo Reviews. A workshop using natural language processing to identify animals from text, group them by species and assess people’s feelings about the animals.
• Spacebound Minds. A workshop with online resources to learn more about mental health and well-being.

Resources from Research Projects

• Risk-averse optimization of genetic circuits - A risk-aware computational framework for synthetic biology that combines Bayesian inference and Thompson sampling to design robust genetic circuits despite model uncertainty. (Kobiela et al., 2026)
• Transforms-For-Brownian-Dynamics - A framework for transforming variable-diffusion Brownian dynamics into constant-diffusion processes via coordinate and time rescaling, enabling more stable and accurate numerical integration. The method supports invertible mappings, improves convergence and efficiency, and extends to multivariate and biophysical diffusion systems. (Phillips et al., 2025)
• Antidepressant MWAS - A large-scale methylome-wide analysis linking antidepressant exposure to whole-blood DNA methylation, integrating self-report and prescription data with validation across external cohorts. The pipeline identifies reproducible CpG associations and builds predictive methylation scores for cross-dataset generalization. (Davyson et al., 2025)
• CheXGenBench - CheXGenBench is a standardized evaluation framework for synthetic chest X-ray generation, assessing fidelity, privacy, and clinical utility across multiple models using a unified multi-metric protocol. It also releases the SynthCheX-75K dataset and provides a reproducible benchmark for comparing and developing medical generative models. (Dutt et al., 2026)
• Uncertainty quantification for WMH segmentation - A framework for evaluating and leveraging uncertainty quantification in WMH segmentation, combining stochastic segmentation networks and deep ensembles to detect failures and ambiguity. It further integrates uncertainty features into downstream Fazekas scoring, improving classification accuracy and calibration. (Philps et al., 2025)
• GNN-Suite - GNN-Suit is a robust modular framework for constructing and benchmarking Graph Neural Network (GNN) architectures in computational biology. GNN-Suite standardises experimentation and reproducibility using the Nextflow workflow to evaluate GNN performance. (Kamp et al., 2025)
• Controlled Amplitude of Present Epitopes (CAPE) - an extension of ProteinMPNN that integrates direct preference optimization (DPO) with MHC-I presentation predictions to design proteins with reduced CTL immunogenicity. The framework preserves structural fidelity while systematically minimizing predicted immune-visible epitopes. (Gasser et al., 2025)
• CrossSDF - a novel method for 3D reconstruction of thin structures from cross-sectional data. It demonstrates a significant improvement over existing methods, effectively reconstructing thin structures and producing accurate 3D models without the interpolation artifacts or over-smoothing of prior approaches. (Walker et al., 2025)
• Diffusion memorisation - a systematic framework for analysing memorization in medical text-to-image diffusion models, identifying high-risk prompts and tokens within MIMIC-CXR. It provides evaluation protocols and mitigation strategies to reduce privacy leakage and improve reliability in synthetic chest X-ray generation. (Dutt, 2025)
• DEPfold - a novel approach to RNA secondary structure prediction that leverages techniques from natural language processing, specifically dependency parsing with biaffine attention. The model can effectively predict both canonical base pairs and pseudoknots, achieving competitive performance on standard RNA structure benchmarks. (Wang & Cohen, 2025)
• VL-ICL Bench - a comprehensive benchmark suite for multimodal in-context learning, encompassing a broad spectrum of tasks that involve both images and text as inputs and outputs, and different types of challenges, from perception to reasoning and long context length. (Zong et al., 2025)
• MOG-Dx - Multi-omic Graph Diagnosis (MOGDx) is a tool for the integration of omic data and classification of heterogeneous diseases. MOGDx exploits a Patient Similarity Network (PSN) framework to integrate omic data using Similarity Network Fusion (SNF). (Ryan et al., 2024)
• BALM - a deep learning framework that leverages pretrained protein and ligand language models to efficiently predict binding affinity, overcoming scalability and generalization limits of existing methods. Paired with improved evaluation strategies, it demonstrates strong performance on unseen targets and practical utility for early-stage drug discovery. (Gorantla et al., 2025)
• Bayesian HRV - A Bayesian model of heart rate variability in bipolar disorder that reveals a probable increase in HRV during recovery from acute episodes, suggesting its potential as a biomarker of symptom resolution. The approach improves tracking of HRV dynamics over time, though effect size estimates remain uncertain due to limited sample size. (Corponi et al., 2024)
• InstanSeg - an efficient, embedding-based instance segmentation pipeline that improves accuracy and reduces processing time for cell and nucleus segmentation in microscopy images. Designed for portability and usability, it supports GPU acceleration, TorchScript deployment, and includes open-source Python and QuPath implementations. (Goldsborough et al., 2024)
• ProbAge - a Python package implementing the probabilistic model for epigenetic age acceleration analysis. This tool provides a mechanistic approach to understanding methylation dynamics and biological aging, addressing limitations in traditional epigenetic clock methods. (Dabrowski et al., 2024)
• CartiMorph Toolbox (CMT) - a deep learning-based pipeline for automated knee cartilage analysis, integrating a novel registration network to quantify cartilage shape and lesions. It delivers competitive performance with a user-friendly framework for imaging biomarker extraction in osteoarthritis. (Yao & Chen, 2024)
• MEDIQA-CORR 2024 - This study evaluates GPT-3.5 and GPT-4 for detecting and correcting medical errors in clinical notes, showing that combining prompting strategies with error-span hints from a smaller model significantly improves performance. The approach achieved a top-10 ranking in MEDIQA-CORR 2024 and highlights key factors affecting LLM reliability in clinical settings. (Gema et al., 2024)
• Self-supervised learning for personal sensing in mood disorders - a self-supervised learning framework for wearable data that overcomes limited annotations to improve detection of acute mood disorder episodes. Using a new dataset and transformer model (E4mer), the approach outperforms supervised methods and highlights the importance of pretraining tasks and unlabeled data scale. (Corponi et al., 2024)
• ICD Coding using ChatGPT - Investigation into GPT-3.5 capabilities in generating and coding medical documents with International Classification of Diseases (ICD)-10 codes for data augmentation on low-resource labels. (Falis et al., 2024)
• Fool your (V)LLMs - This code shows empirically that popular large language and vision-language models are specifically vulnerable to adversarial permutation in answer sets for multiple-choice prompting. These vulnerabilities persist across various model sizes, and exist in very recent language and vision-language models. (Zong et al., 2024)
• VLGuard - A vision-language safe instruction-following dataset that effectively safety aligns VLLMs when integrated this dataset into standard vision-language fine-tuning or utilizing it for post-hoc fine-tuning. The versatility of this safety fine-tuning dataset makes it a valuable resource for safety-testing existing VLLMs, training new models or safeguarding pre-trained VLLMs. (Zong et al., 2024)
• Stochastic Uncertainty Quantification for WMH segmentation - A framework for evaluating stochastic uncertainty quantification methods in WMH segmentation, introducing new metrics (UIRO, JUEO) to analyze inter-analyst variability across datasets. It reveals that differences in annotation policy dominate model uncertainty and proposes task and loss modifications to better align model predictions with diverse labeling strategies. (Philps et al., 2024)
• CAPE-XVAE - a machine learning framework combining protein language models and reinforcement learning to modulate CTL immunogenicity while preserving function. (Gasser et al., 2024)
• Mood disorder symptoms monitoring from multivariate sensory data - A deep learning pipeline for wearable-based monitoring of mood disorders that predicts full HDRS and YMRS symptom profiles rather than single labels. It incorporates multi-task learning and subject-invariant representations, achieving clinically meaningful agreement with expert assessments while addressing key ML challenges. (Corponi et al., 2024)
• GeoGen - an end-to-end generative framework that replaces volumetric density with a learnable signed distance function to produce higher-quality, mesh-consistent 3D geometry from single-view data. By enforcing SDF–depth consistency and using adversarial training, it overcomes NeRF limitations and improves geometric fidelity. (Esposito et al., 2024)
• Active Learning Protocols for Ligand Binding Affinity Prediction - A systematic evaluation of active learning pipelines for drug discovery, comparing models, sampling strategies, and batch sizes across multiple targets. It provides practical guidelines for optimizing binder identification, highlighting advantages of Gaussian processes in low-data regimes and the impact of batch size and noise on performance. (Gorantla et al., 2024)
• FairTune - a bi-level optimisation framework that improves group fairness in medical imaging by selecting parameter-efficient fine-tuning (PEFT) configurations based on validation fairness. It addresses the fairness generalisation gap by balancing model capacity and generalisation, leading to more equitable performance across subgroups. (Dutt et al., 2024)
• Decoding Deep Learning Methods for Binding Affinity Prediction - A systematic analysis of sequence-based deep learning models for binding affinity prediction, probing the impact of protein and ligand representations through architectural comparisons and input perturbations. (Gorantla et al., 2024)
• Parameter-efficient fine-tuning of LLaMA for the clinical domain - A two-stage PEFT framework for clinical NLP that separates domain adaptation and task-specific fine-tuning via specialized adapter layers. It improves performance across clinical prediction tasks, demonstrating efficient adaptation of pretrained language models with reduced computational cost. (Gema et al., 2024)
• PDBench - PDBench is a dataset and software package for evaluating fixed-backbone sequence design algorithms. The structures included in PDBench have been chosen to account for the diversity and quality of observed protein structures, giving a more holistic view of performance. (Castorina et al., 2023)
• Integration of datasets for individual prediction of DNA methylation-based biomarkers - A comparative analysis of DNA methylation normalisation strategies for projecting epigenetic scores across cohorts, evaluating their impact on BMI EpiScore and epigenetic age estimates. It identifies methods that minimise technical variation and ensure consistent individual-level predictions without requiring joint normalisation. (Merzbacher et al., 2023)
• Knowledge graph embeddings in the biomedical domain - An evaluation pipeline for applying and benchmarking state-of-the-art knowledge graph embedding models on the BioKG dataset, achieving substantial gains in link prediction performance. It also integrates rule-based interpretability and validates embeddings on downstream polypharmacy tasks, demonstrating practical utility. (Gema et al., 2023)
• Vaxformer - a conditional protein language model for generating SARS-CoV-2 spike sequences with controllable antigenicity, evaluated via stability, immunogenicity, and structural fidelity metrics. It outperforms CVAE-based approaches, providing a practical framework for computational vaccine design. (Gema et al., 2023)
• MEDFAIR - a comprehensive benchmarking framework for evaluating fairness in medical imaging models, standardizing comparisons across algorithms, datasets, and selection criteria. It highlights the critical impact of model selection on fairness and provides a reproducible pipeline for assessing bias mitigation methods. (Zong et al., 2023)
• DART - a deep learning framework that approximates complex retinal trait extraction pipelines (e.g., fractal dimension) in a single, fast and robust step by training on synthetically degraded images. It enables high-throughput, quality-tolerant inference while closely matching traditional pipeline outputs. (Engelmann et al., 2022)
• Deep attention super-resolution of brain magnetic resonance images - A critic-guided super-resolution framework for enhancing low-quality clinical MRI scans, integrating self-attention and feature-importance for improved interpretability. It produces high-fidelity reconstructions that boost downstream segmentation accuracy, supporting scalable use in research and clinical pipelines. (Li et al., 2022)
• Immune-aware SARS-CoV-2 Spike Proteins for Universal Vaccine Design - A VAE-based generative framework for designing SARS-CoV-2 spike proteins with tunable immune visibility, enabling exploration of vaccine-escape variants. It outperforms simpler generative baselines by producing stable sequences that smoothly interpolate across natural sequence space. (Phillips et al., 2022)
• Detecting multiple retinal diseases in ultra-widefield fundus imaging - A deep learning pipeline for multi-disease detection in ultra-widefield retinal images under realistic clinical conditions, paired with global explainability to localize informative regions. It shows that a small posterior pole region captures most predictive signal, enabling efficient and interpretable diagnosis. (Engelmann et al., 2022)
• CoPHE - Count-preserving hierarchical evaluation and set-based hierarchical evaluation methods for hierarchical label spaces. Currently implemented only for the label space of the ICD-9 ontology of diseases and procedures. (Falis et al., 2021)
• DE-STRESS - a web-based evaluation platform for designed proteins that computes standardized structural metrics and contextual analyses to support data-driven design selection. It streamlines early-stage screening to reduce experimental attrition in protein engineering workflows. (Stam & Wood, 2021)
• ImmunoBERT - a BERT-based model for predicting peptide presentation on MHC Class I molecules, integrating sequence context and multiple MHC alleles. The framework uses SHAP, LIME, and 3D visualizations to interpret predictions, highlighting critical peptide termini and key MHC pockets consistent with biological insights. (Gasser et al., 2021)

  References

1. Kobiela, M., Oyarzún, D. A., & Gutmann, M. U. (2026). Risk-averse optimization of genetic circuits under uncertainty. Cell Systems, 17(1). https://doi.org/10.1016/j.cels.2025.101476
2. Phillips, D., Leimkuhler, B., & Matthews, C. (2025). Numerics with coordinate transforms for efficient Brownian dynamics simulations. Molecular Physics, 123(7-8), e2347546. https://doi.org/10.1080/00268976.2024.2347546
3. Davyson, E., Shen, X., Huider, F., Adams, M. J., Borges, K., McCartney, D. L., Barker, L. F., van Dongen, J., Boomsma, D. I., Weihs, A., Grabe, H. J., Kühn, L., Teumer, A., Völzke, H., Zhu, T., Kaprio, J., Ollikainen, M., David, F. S., Meinert, S., … McIntosh, A. M. (2025). Insights from a methylome-wide association study of antidepressant exposure. Nature Communications, 16(1), 1908. https://doi.org/10.1038/s41467-024-55356-x
4. Dutt, R., Sanchez, P., Yao, Y., McDonagh, S., Tsaftaris, S. A., & Hospedales, T. (2026). CheXGenBench: A Unified Benchmark For Fidelity, Privacy and Utility of Synthetic Chest Radiographs. https://arxiv.org/abs/2505.10496
5. Philps, B., del C. Valdés Hernández, M., Qin, C., Clancy, U., Sakka, E., Muñoz Maniega, S., Bastin, M. E., Jochems, A. C. C., Wardlaw, J. M., & Bernabeu, M. O. (2025). Uncertainty quantification for White Matter Hyperintensity segmentation detects silent failures and improves automated Fazekas quantification. Medical Image Analysis, 105, 103697. https://doi.org/https://doi.org/10.1016/j.media.2025.103697
6. Kamp, S., Stracquadanio, G., & Simpson, T. I. (2025). GNN-Suite: a Graph Neural Network Benchmarking Framework for Biomedical Informatics. ArXiv Preprint ArXiv:2505.10711. https://doi.org/10.48550/arXiv.2505.10711
7. Gasser, H.-C., Oyarzún, D. A., Alfaro, J. A., & Rajan, A. (2025). Tuning ProteinMPNN to reduce protein visibility via MHC Class I through direct preference optimization. Protein Engineering, Design and Selection, 38, gzaf003. https://doi.org/10.1093/protein/gzaf003
8. Walker, T., Esposito, S., Rebain, D., Vaxman, A., Onken, A., Li, C., & Mac Aodha, O. (2025). CrossSDF: 3D Reconstruction of Thin Structures From Cross-Sections. Proceedings of the Computer Vision and Pattern Recognition Conference (CVPR), 30928–30937.
9. Dutt, R. (2025). The Devil is in the Prompts: De-Identification Traces Enhance Memorization Risks in Synthetic Chest X-Ray Generation. https://arxiv.org/abs/2502.07516
10. Wang, K., & Cohen, S. B. (2025). DEPfold: RNA Secondary Structure Prediction as Dependency Parsing. International Conference on Representation Learning, 2025, 9800–9821. https://proceedings.iclr.cc/paper_files/paper/2025/file/1ad84bf5a6711cb9541c8976617cc00f-Paper-Conference.pdf
11. Zong, Y., Bohdal, O., & Hospedales, T. (2025). VL-ICL Bench: The Devil in the Details of Multimodal In-Context Learning. https://arxiv.org/abs/2403.13164
12. Ryan, B., Marioni, R. E., & Simpson, T. I. (2024). Multi-Omic Graph Diagnosis (MOGDx): a data integration tool to perform classification tasks for heterogeneous diseases. Bioinformatics, 40(9), btae523. https://doi.org/10.1093/bioinformatics/btae523
13. Gorantla, R., Gema, A. P., Yang, I. X., Serrano-Morrás, Á., Suutari, B., Juárez-Jiménez, J., & Mey, A. S. J. S. (2025). Learning Binding Affinities via Fine-Tuning of Protein and Ligand Language Models. Journal of Chemical Information and Modeling, 65(22), 12279–12291. https://doi.org/10.1021/acs.jcim.5c02063
14. Corponi, F., Li, B. M., Anmella, G., Valenzuela-Pascual, C., Pacchiarotti, I., Valentí, M., Grande, I., Benabarre, A., Garriga, M., Vieta, E., Lawrie, S. M., Whalley, H. C., Hidalgo-Mazzei, D., & Vergari, A. (2024). A Bayesian analysis of heart rate variability changes over acute episodes of bipolar disorder. Npj Mental Health Research, 3(1), 44. https://doi.org/10.1038/s44184-024-00090-x
15. Goldsborough, T., Philps, B., O’Callaghan, A., Inglis, F., Leplat, L., Filby, A., Bilen, H., & Bankhead, P. (2024). InstanSeg: an embedding-based instance segmentation algorithm optimized for accurate, efficient and portable cell segmentation. https://arxiv.org/abs/2408.15954
16. Dabrowski, J. K., Yang, E. J., Crofts, S. J. C., Hillary, R. F., Simpson, D. J., McCartney, D. L., Marioni, R. E., Kirschner, K., Latorre-Crespo, E., & Chandra, T. (2024). Probabilistic inference of epigenetic age acceleration from cellular dynamics. Nature Aging, 4(10), 1493–1507. https://doi.org/10.1038/s43587-024-00700-5
17. Yao, Y., & Chen, W. (2024). Quantifying Knee Cartilage Shape and Lesion: From Image to Metrics. https://arxiv.org/abs/2409.07361
18. Gema, A. P., Lee, C., Minervini, P., Daines, L., Simpson, T. I., & Alex, B. (2024). Edinburgh Clinical NLP at MEDIQA-CORR 2024: Guiding Large Language Models with Hints. https://arxiv.org/abs/2405.18028
19. Corponi, F., Li, B. M., Anmella, G., Valenzuela-Pascual, C., Mas, A., Pacchiarotti, I., Valentí, M., Grande, I., Benabarre, A., Garriga, M., Vieta, E., Young, A. H., Lawrie, S. M., Whalley, H. C., Hidalgo-Mazzei, D., & Vergari, A. (2024). Wearable Data From Subjects Playing Super Mario, Taking University Exams, or Performing Physical Exercise Help Detect Acute Mood Disorder Episodes via Self-Supervised Learning: Prospective, Exploratory, Observational Study. JMIR Mhealth Uhealth, 12, e55094. https://doi.org/10.2196/55094
20. Falis, M., Gema, A. P., Dong, H., Daines, L., Basetti, S., Holder, M., Penfold, R. S., Birch, A., & Alex, B. (2024). Can GPT-3.5 generate and code discharge summaries? Journal of the American Medical Informatics Association, 31(10), 2284–2293. https://doi.org/10.1093/jamia/ocae132
21. Zong, Y., Yu, T., Chavhan, R., Zhao, B., & Hospedales, T. (2024). Fool Your (Vision and) Language Model With Embarrassingly Simple Permutations. https://arxiv.org/abs/2310.01651
22. Zong, Y., Bohdal, O., Yu, T., Yang, Y., & Hospedales, T. (2024). Safety Fine-Tuning at (Almost) No Cost: A Baseline for Vision Large Language Models. https://arxiv.org/abs/2402.02207
23. Philps, B., del C. Valdes Hernandez, M., Munoz Maniega, S., Bastin, M. E., Sakka, E., Clancy, U., Wardlaw, J. M., & Bernabeu, M. O. (2024). Stochastic Uncertainty Quantification Techniques Fail to Account for Inter-analyst Variability in White Matter Hyperintensity Segmentation. Medical Image Understanding and Analysis, 34–53.
24. Gasser, H.-C., Oyarzún, D. A., Rajan, A., & Alfaro, J. A. (2024). Guiding a language-model based protein design method towards MHC Class-I immune-visibility targets in vaccines and therapeutics. ImmunoInformatics, 14, 100035. https://doi.org/https://doi.org/10.1016/j.immuno.2024.100035
25. Esposito, S., Xu, Q., Kania, K., Hewitt, C., Mariotti, O., Petikam, L., Valentin, J., Onken, A., & Mac Aodha, O. (2024). GeoGen: Geometry-Aware Generative Modeling via Signed Distance Functions. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) Workshops, 7479–7488.
26. Gorantla, R., Kubincová, A., Suutari, B., Cossins, B. P., & Mey, A. S. J. S. (2024). Benchmarking Active Learning Protocols for Ligand-Binding Affinity Prediction. Journal of Chemical Information and Modeling, 64(6), 1955–1965. https://doi.org/10.1021/acs.jcim.4c00220
27. Dutt, R., Bohdal, O., Tsaftaris, S. A., & Hospedales, T. (2024). FairTune: Optimizing Parameter Efficient Fine Tuning for Fairness in Medical Image Analysis. https://arxiv.org/abs/2310.05055
28. Gema, A. P., Minervini, P., Daines, L., Hope, T., & Alex, B. (2024). Parameter-Efficient Fine-Tuning of LLaMA for the Clinical Domain. https://arxiv.org/abs/2307.03042
29. Castorina, L. V., Petrenas, R., Subr, K., & Wood, C. W. (2023). PDBench: Evaluating Computational Methods for Protein-Sequence Design. Bioinformatics. https://doi.org/10.1093/bioinformatics/btad027
30. Merzbacher, C., Ryan, B., Goldsborough, T., Hillary, R. F., Campbell, A., Murphy, L., McIntosh, A. M., Liewald, D., Harris, S. E., McRae, A. F., Cox, S. R., Cannings, T. I., Vallejos, C. A., McCartney, D. L., & Marioni, R. E. (2023). Integration of datasets for individual prediction of DNA methylation-based biomarkers. Genome Biology, 24(1), 278. https://doi.org/10.1186/s13059-023-03114-5
31. Gema, A. P., Grabarczyk, D., Wulf, W. D., Borole, P., Alfaro, J. A., Minervini, P., Vergari, A., & Rajan, A. (2023). Knowledge Graph Embeddings in the Biomedical Domain: Are They Useful? A Look at Link Prediction, Rule Learning, and Downstream Polypharmacy Tasks. https://arxiv.org/abs/2305.19979
32. Gema, A. P., Kobiela, M., Fraisse, A., Rajan, A., Oyarzún, D. A., & Alfaro, J. A. (2023). Vaxformer: Antigenicity-controlled Transformer for Vaccine Design Against SARS-CoV-2. https://arxiv.org/abs/2305.11194
33. Zong, Y., Yang, Y., & Hospedales, T. (2023). MEDFAIR: Benchmarking Fairness for Medical Imaging. https://arxiv.org/abs/2210.01725
34. Engelmann, J., Villaplana-Velasco, A., Storkey, A., & Bernabeu, M. O. (2022). Robust and Efficient Computation of Retinal Fractal Dimension Through Deep Approximation. Ophthalmic Medical Image Analysis, 84–93.
35. Li, B. M., Castorina, L. V., Valdés Hernández, M. del C., Clancy, U., Wiseman, S. J., Sakka, E., Storkey, A. J., Jaime Garcia, D., Cheng, Y., Doubal, F., Thrippleton, M. T., Stringer, M., & Wardlaw, J. M. (2022). Deep attention super-resolution of brain magnetic resonance images acquired under clinical protocols. Frontiers in Computational Neuroscience, Volume 16 - 2022. https://doi.org/10.3389/fncom.2022.887633
36. Phillips, D., Gasser, H.-C., Kamp, S., Pałkowski, A., Rabalski, L., Oyarzún, D. A., Rajan, A., & Alfaro, J. A. (2022). Generating Immune-aware SARS-CoV-2 Spike Proteins for Universal Vaccine Design. Proceedings of the 1st Workshop on Healthcare AI and COVID-19, ICML 2022, 184, 100–116. https://proceedings.mlr.press/v184/phillips22a.html
37. Engelmann, J., McTrusty, A. D., MacCormick, I. J. C., Pead, E., Storkey, A., & Bernabeu, M. O. (2022). Detecting multiple retinal diseases in ultra-widefield fundus imaging and data-driven identification of informative regions with deep learning. Nature Machine Intelligence, 4(12), 1143–1154. https://doi.org/10.1038/s42256-022-00566-5
38. Falis, M., Dong, H., Birch, A., & Alex, B. (2021). CoPHE: A Count-Preserving Hierarchical Evaluation Metric in Large-Scale Multi-Label Text Classification. Proceedings of the 2021 Conference on Empirical Methods in Natural Language Processing, 907–912. https://doi.org/10.18653/v1/2021.emnlp-main.69
39. Stam, M. J., & Wood, C. W. (2021). DE-STRESS: a user-friendly web application for the evaluation of protein designs. Protein Engineering, Design and Selection, 34, gzab029. https://doi.org/10.1093/protein/gzab029
40. Gasser, H.-C., Bedran, G., Ren, B., Goodlett, D., Alfaro, J., & Rajan, A. (2021). Interpreting BERT architecture predictions for peptide presentation by MHC class I proteins. https://arxiv.org/abs/2111.07137

---