Gut microbes track deep sleep, not total sleep.

Prior cohorts (typically n < 30) rely on consumer-grade actigraphy, which is blind to sleep stage — it cannot tell slow-wave sleep from light sleep. None have paired 16S-or-deeper microbiome profiling with polysomnography-grade staging in the same participants, and few control for the confounders that move both gut and sleep at once.

• Actigraphy-only designs can report "sleep quality" but not sleep architecture.
• Diet, BMI, and antibiotic history are rarely logged as covariates.
• No prior cohort repeated sampling across a full six-week window.

Stool samples were collected at week 0 and week 6 and run through shotgun metagenomic sequencing. All 64 participants wore actigraphy nightly; a 22-person subsample also underwent overnight polysomnography at both timepoints, giving us gold-standard sleep staging alongside microbial composition.

Diet was logged by 3-day food record each week. Linear mixed-effects models carry participant as a random intercept and adjust for BMI, dietary fiber intake, and antibiotic use in the prior 90 days.

• Week 0 & week 6 stool sampling, shotgun metagenomics (Illumina NovaSeq).
• Nightly actigraphy, all 64; overnight PSG on a 22-person subsample.
• LME models adjust for BMI, fiber intake, and 90-day antibiotic history.