Most models are still trained on methylation data, all associations are true only at the population level. Great metric for global health/policy making/cohort-based analytics.

But, for an individual, due to high variance throughout the day, we can not conclude much out of this at its face value without adding a huge uncertainty window. Terrible for personalised decision making on supplements, etc