Can we distinguish biological correlation from true causation in our DNA?
In short: In genomics, seeing two genes change together doesn't mean one drives the other. To move beyond mere patterns, scientists are now using causal machine learning and massive datasets to untangle the complex regulatory networks that govern life and development.
For decades, genomics has relied on identifying correlations within 'omics' data—genomics, transcriptomics, and epigenomics—to understand biological processes. However, a persistent challenge remains: differentiating between simple associations and true causal mechanisms. As highlighted in MIT's 7.91J course by Professor George Church, the goal is to use genomic data and natural computing tools to bridge this gap.
One frontier involves studying de novo mutations (DNMs) in the germline, which are linked to various genetic disorders. By leveraging the UK's 100,000 Genomes Project, which includes over 10,000 parent-offspring trios, researchers have gained unprecedented statistical power to investigate these mutations. While recent studies have not found statistically significant evidence for the non-zero heritability of DNM counts, they have confirmed that environmental factors play a significant role in their causality. One example involves researchers linking premature menopause to higher DNM levels by analysing UK Biobank datasets via Mendelian randomisation.
Modern computational approaches, such as deep graph neural networks, are now being applied to single-cell RNA-seq data to infer gene regulatory networks (GRNs) during developmental processes like gastrulation. Despite these advances, causal machine learning in single-cell genomics faces three major hurdles: the difficulty of learning cell dynamics, the complexity of interpreting learned models, and the lack of model generalization to entirely new experimental conditions. Overcoming these challenges is essential to truly understanding how targeted perturbations affect the whole transcriptome and drive cellular mechanisms.
Opens the Intellect+ feed with this card first. Swipe up for the next one.
Sources
- Summary by Intellect · source: 22. Causality, Natural Computing, and Engineering Genomes
- Also: Causal inference in integrative genomics: exploring de novo mutations and unravelling causal mechanisms via gene regulatory networks in developmental process
- Also: Causal machine learning for single-cell genomics | Nature Genetics
To cite this card, link to https://intellectplus.co.uk/app/ideas-22-causality-natural-computing-and-engineering-genomes.
How we source, check and correct cards: Editorial standards.
General information, not professional advice. Spotted an error? Flag the card or see corrections and complaints.
Intellect Plus is a visual educational discovery platform for exploring fascinating facts and ideas. Brain Food, Not Brain Rot. What is Intellect Plus?
