Experimental design for high dimensional data (including omics)

This training presentation introduces key principles for designing robust, reproducible studies involving high-dimensional biological data, including omics and other complex biomedical datasets. It explains why early experimental design decisions are critical for separating true biological signal from noise, avoiding confounding, and ensuring that downstream statistical and machine-learning analyses are valid. The session covers common sources of bias and variability, including batch effects, site effects, inconsistent sample processing, insufficient metadata, pseudoreplication, poor randomisation, and inappropriate handling of missing values or normalisation. Through practical biomedical examples, the presentation highlights how poor design can lead to misleading results, wasted resources, and data that cannot address the primary research question. The presentation also provides practical recommendations for improving study quality, including balanced sample processing, randomisation within batches, appropriate use of biological replicates, inclusion of reference and quality-control samples, careful metadata capture, regular data audits, transparent analysis reporting, and adherence to FAIR and Open Research principles. It emphasises the importance of interdisciplinary collaboration between clinicians, laboratory scientists, statisticians, bioinformaticians, data stewards and computational biologists from the earliest stages of study planning. This resource is intended for researchers, clinicians, postgraduate students, technical professionals and research support staff involved in the design, generation, analysis or interpretation of complex biomedical datasets. This is a presentation part of the work package Nurture from the MRC-funded BIOMEDASA, led by the University of Liverpool. Nuture aims to propagate best research practices among non-specialists with short accessible training. Topics spanning study design, FAIR/Open Science and Introduction to GenAI and responsible use in research. This particular presentation covers a short introduction to FAIR/Open Science. Our main aim is to build a stronger Open Research culture, higher data quality, and improved reproducibility. We are very happy to run sessions on these topics in any institution so please reach out to biomedasaatliverpool.ac.uk if you wish to engage further.