TL;DR
By mining 1.5 million primary care records from the UK Biobank, researchers have mapped the genetic "blueprint" of how our weight changes over time, rather than just looking at a single snapshot. They identified six novel genetic regions associated with weight trajectories—including a variant in the APOE gene linked to weight loss—and discovered that the genes driving weight gain are largely distinct from those determining your starting weight.
What Problem Does This Paper Address?
Most of what we know about the genetics of obesity comes from "cross-sectional" studies—essentially taking a single measurement of Body Mass Index (BMI) from thousands of people and looking for correlations. While this has identified over 1,000 genetic markers, it treats weight as a static trait.
In reality, adiposity (body fat) is dynamic. Some people gain weight steadily throughout adulthood, while others remain stable or lose weight. The "gap" the authors address is our lack of understanding regarding the genetic contribution to these longitudinal trajectories. By understanding the genetics of change, we might uncover the biological "engines" of weight gain and loss, which could lead to more personalized prevention and treatment strategies.
Methods
The researchers developed a sophisticated two-stage analytical pipeline to turn messy Electronic Health Record (EHR) data into clean genetic phenotypes:
- Phenotype Derivation: They used two modeling approaches. First, Linear Mixed-Effects (LME) models captured the "slope" of weight change for each individual. Second, they used Regularized B-splines to model non-linear changes (curves).
- Soft Clustering: Using the B-spline coefficients, they performed k-medoids clustering to group individuals into four trajectory patterns: high gain, moderate gain, stable, and loss. They used "soft" clustering, meaning they calculated the probability of an individual belonging to a specific group.
- GWAS: They then performed Genome-Wide Association Studies (GWAS) on these derived traits (slopes and cluster probabilities), adjusting for baseline BMI to ensure they were finding genes specifically related to change.
Datasets
The study leveraged the UK Biobank (UKBB), specifically focusing on a subset of ~177,000 individuals of white British ancestry. The "secret sauce" was linking these participants to their longitudinal primary care EHRs, totaling 1.5 million BMI and weight measurements spanning several decades.
For validation, they used a hold-out set of ~17,000 UKBB participants with repeat measurements taken at assessment centers, as well as self-reported weight change data from over 300,000 participants.
Key Insights and Ideas
- Repeat Measures Boost Power: Even for "baseline" BMI, using the average of multiple measurements from a health record is statistically more powerful than a single snapshot. The team identified 14 novel loci for baseline BMI simply by using this longitudinal "noise reduction."
- The APOE Connection: The most striking find was rs429358, a missense variant in the APOE gene. While famous for its link to Alzheimer’s and lipid metabolism, this study shows it is robustly associated with weight loss over time, independent of a person's starting weight.
- Low Genetic Overlap: The genetic correlation between baseline BMI and BMI-change was only about 45%. This is a major "aha" moment: it suggests that the biology of being heavy is significantly different from the biology of becoming heavy.
- Sex Matters in Trajectories: The heritability of weight change was significantly higher in women (2.82%) than in men (1.38%). This suggests that genetic factors may play a larger role in how women’s weight fluctuates through adulthood (potentially influenced by hormonal shifts like menopause).
Key Results
- Novel Loci: Identified 6 independent genetic loci associated with adult adiposity change.
- Heritability: The SNP-based heritability of BMI-change was calculated at 1.98%. While this is 9-fold lower than the heritability of static BMI (~25-30% in this cohort), it confirms a clear, measurable genetic component to weight trajectories.
- Biomarker Links: The APOE variant associated with weight loss was also linked to longitudinal changes in other health markers, including increases in total cholesterol and C-reactive protein (a marker of inflammation), but decreases in triglycerides.
Limitations and Future Work
The authors acknowledge several hurdles:
- Ancestry Bias: The study was limited to individuals of white British ancestry. Because genetic architecture varies across populations, these results may not generalize to other ethnic groups.
- EHR Noise: Primary care data is "found" data, not "collected" data. It contains inaccuracies, and people are more likely to have their weight recorded when they are ill (informed presence bias).
- Causality: It remains unclear if the identified genes cause weight change or are associated with behaviors or medications (like statins) that influence weight.
Future Work: The researchers suggest that as biobanks grow and integrate more diverse EHR data, we can move toward "longitudinal phenome-wide scans" to see how genetic risks for weight gain interact with other biomarkers and disease outcomes over a lifetime.
Summary updated