Computational methods for multi-modal data analysis

IIDS67692 Computational methods for multi-modal data analysis 15 Semester 2 Postgraduate Taught Level 7 Mobarak Hoque Unit coordinator Lijing Lin Unit coordinator Sokratia Georgaka Unit coordinator Division of Informatics, Imaging and Data Sciences 1 FHEQ level (Framework for Higher Education Qualifications) ' Masters/Integrated Masters P4 ' European Credit Transfer & Accumulation System Rating : 7.5 The analysis of multi-modal data has significantly advanced our understanding of complex biological systems and their applications in healthcare. By integrating omics, imaging, and health data using cutting-edge machine learning methods, researchers can infer relationships across diverge modalities. This unit equips students with fundamental knowledge and hands-on computational skills to analyse and interpret such multi-modal datasets effectively. Three main areas will be covered to give students the range of skills to be able to work with large scale, multi-modal data:1) Introduction to methods for omics, health data and imaging data analysis<ul><li>What is omics data? Introduction to key terminologies. (e.g. multiomics/multiome, single-cell, spatial transcriptomics, spatial proteomics, imaging etc) and health data recap</li><li>Why we need this kind of data?</li><li>From bulk to single-cell – example pipeline of computational analysis e.g. scanpy</li><li>Examples of research studies/published literature</li></ul>2) Introduction to multi-modal data,<ul><li>advantages and disadvantages of multi-modal data analyses/research</li><li>cases studies of multi-modal health research projects</li></ul>3) Computational methods for analysing multi-modal omics data<ul><li>Data integration techniques</li><li>Spatial omics data practical exercise</li><li>Statistical modelling and inference</li><li>Tools and pipelines</li><li>Applications and case studies e.g. LLMs/genAI</li></ul>  The analysis of multi-modal data has significantly advanced our understanding of complex biological systems and their applications in healthcare. By integrating omics, imaging, and health data using cutting-edge machine learning methods, researchers can infer relationships across diverge modalities. This unit equips students with fundamental knowledge and hands-on computational skills to analyse and interpret such multi-modal datasets effectively.Three main areas will be covered to give students the range of skills to be able to work with large scale, multi-modal data:1) Introduction to methods for omics, health data and imaging data analysis<ul><li>What is omics data? Introduction to key terminologies. (e.g. multiomics/multiome, single-cell, spatial transcriptomics, spatial proteomics, imaging etc) and health data recap</li><li>Why we need this kind of data?</li><li>From bulk to single-cell – example pipeline of computational analysis e.g. scanpy</li><li>Examples of research studies/published literature</li></ul>2) Introduction to multi-modal data,<ul><li>advantages and disadvantages of multi-modal data analyses/research</li><li>case studies of multi-modal health research projects</li></ul>3) Computational methods for analysing multi-modal omics data<ul><li>Data integration techniques</li><li>Spatial omics data practical exercise</li><li>Statistical modelling and inference</li><li>Tools and pipelines</li><li>Applications and case studies e.g. LLMs/genAI</li></ul> The unit aims to:Introduce the concept of multi-modal data for health and computational techniques applicable to investigate diverse data types. Different case studies will be presented allowing the learners to explore the breadth of this topic and emerging research areas. It is also designed to prepare students for multidisciplinary research by integrating concepts and techniques from multiple domains.  Students will be able to:<ol><li>Describe the principles of multi-modal omics, its applications in integrating diverse datasets, and its role in emerging research areas.</li><li>Demonstrate knowledge of computational techniques (ML/AI) and tools, for analysing multi-modal omics data.</li><li>Explain the challenges and limitations of integrating multi-modal omics data.</li><li>Explain the data types and data sources involved in examples of multi-modal data health research </li></ol> <ol><li>Critically analyse multi-modal omics/health data and compare different computational methods</li><li>Evaluate scientific and clinical research literature to understand the current research landscape</li><li>Interpret findings from diverse data technologies </li></ol> <ol><li>Apply machine-learning approaches for multi-modal data analyses</li><li>Conduct data analyses on example datasets</li><li>Develop workflows for multi-modal data analysis, employing methods such as data integration, modelling, and visualisation. </li></ol> <ol><li>Collaborate in teams that bring together expertise from computational, experimental, and clinical domains.</li><li>Demonstrate independent problem-solving skills and adaptability in handling diverse and complex datasets.</li><li>Gain knowledge and hands on experience on ML approaches for integrating multi-modal data which is transferable to other domains.</li><li>Enhance verbal and written communication skills when appraising and assessing research literature in groups. </li></ol> This module will be delivered in a ‘blended’ learning format over ~6 weeks. Teaching material, including lectures that will be available online either prior to face-to-face (F2F) teaching sessions/tutorials or after live/F2F delivery. Group activities (formative assessments with feedback from tutors and peers) will be facilitated during F2F sessions to consider and reflect on a range of topics/case studies covered in the module.  <figure class="table"><table><tbody><tr><td>Assessment Task</td><td>How and when feedback is provided</td><td>Weighting within unit (if relevant)</td></tr><tr><td>Formative assessment</td><td>During F2F workshops/groupwork or via online progress checkpoints (e.g. short quizzes) </td><td>NA</td></tr><tr><td>Summative assessment:Practical programming exercise using provided input data. This will involve using Jupyter notebooks to run multi-modal analysis pipeline on the provided input data. This can, for example, include assessing the effect of the model parameters to the output or comparing the outcome of two different short pipelines and to write a report that summarises the results.  This task will be given at the end of the module and the students will be given three weeks to complete this task. Learners submit their notebook that details how they conduct their analysis.  </td><td>Written feedback provided within 15 days of assignment submission deadline. </td><td>100%</td></tr></tbody></table></figure> Written feedback provided within 15 days of assignment submission deadline N <ul><li>Spatial omics technologies at multimodal and single cell/subcellular level: <a href="https://genomebiology.biomedcentral.com/articles/10.1186/s13059-022-02824-6">https://genomebiology.biomedcentral.com/articles/10.1186/s13059-022-02824-6</a></li><li>Single-cell multimodal omics: the power of many: <a href="https://www.nature.com/articles/s41592-019-0691-5">https://www.nature.com/articles/s41592-019-0691-5</a></li><li>Reccomended Methods and applications for single-cell and spatial multi-omics: <a href="https://www.nature.com/articles/s41576-023-00580-2">https://www.nature.com/articles/s41576-023-00580-2</a></li><li>The dawn of spatial omics:<a href="https://pmc.ncbi.nlm.nih.gov/articles/PMC7614974/">https://pmc.ncbi.nlm.nih.gov/articles/PMC7614974/</a></li><li>Single-cell multimodal omics: the power of many <a href="https://www.nature.com/articles/s41592-019-0691-5">https://www.nature.com/articles/s41592-019-0691-5</a></li><li>MUON: multimodal omics analysis framework <a href="https://genomebiology.biomedcentral.com/articles/10.1186/s13059-021-02577-8">https://genomebiology.biomedcentral.com/articles/10.1186/s13059-021-02577-8</a></li><li>An approach for integrating multimodal omics data into sparse and interpretable models<a href="https://pmc.ncbi.nlm.nih.gov/articles/PMC10921032/ ">https://pmc.ncbi.nlm.nih.gov/articles/PMC10921032/ </a></li></ul> Lectures 17 Practical classes & workshops 18 0 115