Designing Fiber–Gut Microbiome Interactions with Active Learning
Most prebiotic research starts with an ingredient and tests what it does. This study applied a different approach where a team at Duke University and the University of Washington built a system that predicts which fiber and bacteria combinations will produce a specific gut outcome, then builds and tests those exact combinations to see if the prediction holds.
The researchers combined machine learning, Bayesian optimization, and robotic automation to run a design-test-learn cycle: predict a promising fiber-species pairing, build it on the bench, measure what it actually does, and feed those results back into the model before the next round. They ran five rounds this way, testing 15 gut bacteria species against 6 dietary fibers, with robots handling up to 390 conditions at once. Even with just 21 variables on the table, the number of possible combinations ran into the trillions, far more than any lab could test by hand.
The model landed on a specific, repeatable combination. Inulin paired with two bacterial species, Bacteroides uniformis and Anaerostipes caccae, consistently produced high, stable levels of butyrate, the short-chain fatty acid tied to gut and metabolic health. A third species, Prevotella copri, added a further boost when present. And when the researchers dropped this exact combination into real human fecal samples instead of lab-built communities, it produced the same beneficial result the model had predicted.
This is a mechanistic lab-based study, not a human clinical trial, but it points toward where prebiotic science is heading: matching specific fibers to specific microbial partners for a particular outcome, providing evidence that “fiber” should not be thought of as one interchangeable ingredient. As the researchers put it, companies are already combining bacteria and fibers in products, but few are matching the right synergy to the right problem yet.
Key Takeaways:
- Researchers used machine learning, Bayesian optimization, and robotic automation to systematically test which fiber and gut bacteria combinations reliably produce beneficial outcomes, instead of testing one ingredient at a time.
- The model identified a specific combination, inulin plus Bacteroides uniformis and Anaerostipes caccae, that consistently produced high butyrate levels, and this held up when tested in real human fecal samples.
- This is lab-based, mechanistic work rather than a clinical trial, but it offers a genuinely new framework for designing precision fiber-microbiome pairings.
Access the study: https://doi.org/10.1038/s41589-026-02272-4
Reference: Connors, B. M., Thompson, J., Gangan, M. S., Quinn-Bohmann, N., Gibbons, S. M., Grant, J., Castellanos-Sanchez, A., McCann, J. R., Rawls, J. F., & Venturelli, O. S. (2026). Designing fiber-gut microbiome interactions with active learning. Nature Chemical Biology, 22(8), 1286–1298. https://doi.org/10.1038/s41589-026-02272-4
Safety, Tolerability, and Gastrointestinal Effect of a Bifidobacterium-Based Synergistic Synbiotic in Infants and Toddlers (The ARTEMIS Trial)
In a healthy infant gut, Bifidobacterium species are typically found to be dominant. However, in developed countries, infant microbiomes are often less diverse than is considered ideal. This has led researchers to investigate targeted synbiotics as opposed to single-strain probiotics to address this need.
The ARTEMIS trial, tested a synbiotic combining four proprietary Bifidobacterium strains with a blend of human milk oligosaccharides (HMOs) and vitamin D in 114 infants (2–12 months) and toddlers (12–24 months), randomized to synbiotic or placebo for 4 weeks with a 2-week washout. Safety and tolerability were the primary endpoints, and they found that adverse events and GI symptoms didn’t differ significantly between the synbiotic and placebo groups in either age cohort.
The secondary data is also significant. After 4 weeks, at least one administered Bifidobacterium strain showed up in 72% of infants and 67% of toddlers, and engraftment for two of the four strains persisted through the full 2-week washout, notable because typical probiotic strains like Lactobacillus are usually transient and clear out once dosing stops. Supplementation also significantly increased Bifidobacterium infantis abundance and HMO-utilization genes, and toddlers on the synbiotic reported significantly improved sleep quality, an exploratory finding future trials should test directly.
Key Takeaways:
- This is the first U.S. clinical trial of a synergistic Bifidobacterium + HMO synbiotic in infants and toddlers, and it found the product safe and well tolerated, with no significant difference in adverse events versus placebo.
- Supplementation significantly increased both Bifidobacterium infantis abundance and HMO-utilization genes together, evidence that the HMOs in the blend were actively feeding the probiotic strains.
- Toddlers on the synbiotic also reported improved sleep, an exploratory finding that opens a new angle for future trials.
Access the study: https://doi.org/10.1038/s41390-026-05318-4
Reference: Jarman, J. B., Torres, P. J., Baum, C., Tinoco, J., Sato, H., Rasteiro, C. S., Selbrede, R., Insel, R., Culler, S. J., & Van Dien, S. (2026). Safety, tolerability, and gastrointestinal effect of a Bifidobacterium-based synergistic synbiotic in infants and toddlers. Pediatric Research. https://doi.org/10.1038/s41390-026-05318-4
Fiber Supplementation in Irritable Bowel Syndrome: A Systematic Review and Meta-Analysis of Randomized Controlled Trials
This meta-analysis pools 30 randomized controlled trials and 1,904 adults with IBS across four fiber categories: wheat bran, psyllium, inulin-type fructans, and B-GOS. That’s the largest and most current evidence base assembled on fiber in IBS, and instead of pooling everything into one “fiber” effect or the older soluble-versus-insoluble split, the authors run subgroup analyses by individual fiber type (wheat bran, psyllium, inulin-type fructans, and B-GOS), citing recent evidence that fermentability and viscosity, not just solubility, are what actually determine how a fiber behaves in the gut.
Fiber overall improved clinical response, with 52% of participants responding versus 44% on control. Almost all of that came from psyllium (RR 1.53 [95% CI 1.06-2.19], P=0.02, across 5 trials); wheat bran and inulin-type fructans didn’t move responder rates on their own. Psyllium gets there through viscosity and water-holding capacity. It’s a soluble, low-fermentability fiber that forms a viscoelastic gel in the gut, that made it popular as an over-the-counter laxative.
Another significant finding was with B-GOS. Across all four categories, the pooled effect on overall GI symptoms wasn’t significant, but GOS’s subgroup at doses up to 7.0 g/day was significant. This result comes from 2 trials and 64 total participants. Glucomannan also showed a significant effect on the same overall symptom measure, with a larger point estimate, but this is based on a single 15-person trial.
This study reinforces the fact that fiber types don’t behave alike. The value here is the scale and precision behind that idea, with 30 trials and nearly 1,900 patients. The dosing guidance is usable, too: up to 30 g/day of psyllium for clinical response and up to 7.0 g/day of B-GOS for overall symptom relief.
Key Takeaways:
- A meta-analysis of 30 RCTs and 1,904 adults with IBS found fiber overall improved clinical response (52% vs. 44% on control), however wheat bran and inulin-type fructans showed no effect on response rates.
- B-GOS significantly improved overall GI symptoms at doses up to 7.0 g/day, though this result was seen using data from only 2 trials and 64 participants. Glucomannan also hit significance on the same measure with a larger effect size, from a single 15-person trial.
Access the study: https://doi.org/10.1053/j.gastro.2026.07.027
Reference: Staudacher, H. M., Rucco, V., Mancell, S., Dimidi, E., & Whelan, K. (2026). Fiber supplementation in irritable bowel syndrome: a systematic review and meta-analysis of randomized controlled trials. Gastroenterology. https://doi.org/10.1053/j.gastro.2026.07.027
Improvement Effect of Synbiotics Intake on Stool Frequency in Constipation-Prone Healthy Participants
Chronic constipation is common, and while probiotics and prebiotics have each shown some benefit for defecation symptoms on their own, whether combining them adds anything is still an open question.
This placebo-controlled, randomized, double-blind trial tested a fermented milk beverage combining Lacticaseibacillus paracasei strain Shirota (LcS) and galacto-oligosaccharides (GOS), dosed at 40 billion CFU LcS plus 5 g GOS daily, in 200 healthy, constipation-prone Japanese adults over 4 weeks.
Defecation frequency was significantly higher in the synbiotic group at 2 weeks, but that effect didn’t hold up by week 4, a limitation the authors acknowledge directly. Other measures fared better, with defecation volume and stool consistency both showing significant improvement. It was also shown that fecal Bifidobacterium counts rose, and putrefactive compounds linked to bacterial breakdown dropped. But because the placebo didn’t isolate for either LcS and GOS, the trial can’t say which component, or the combination, led to these improvements.
Key Takeaways:
- A 200-person RCT tested a synbiotic combining LcS (a probiotic strain) and GOS (a prebiotic) in constipation-prone adults, finding that defecation frequency rose significantly at 2 weeks but didn’t hold at 4 weeks.
- Stool consistency and defecation volume improved significantly, and fecal Bifidobacterium counts rose alongside a drop in putrefactive compounds.
- The placebo excluded both study components, so the trial can’t isolate whether GOS, LcS, or their combination drove the benefits, and all study authors are Yakult employees.
Access the study: https://doi.org/10.1163/18762891-BJA00134
Reference: Shima, T., Kaga, C., Kado, Y., Date, R., Shimamoto, K., Tajima, S., Makino, H., Shimizu, K., Katto, M., Kato, A., Tsuji, H., & Matsumoto, S. (2026). Improvement effect of synbiotics intake on stool frequency in constipation-prone healthy participants: placebo-controlled, randomised, double-blind, parallel-group comparison. Beneficial Microbes. https://doi.org/10.1163/18762891-BJA00134
