Today's Peptide News — August 8, 2026

David Baker's lab has extended its deep-learning toolkit — RoseTTAFold2 and RFdiffusion — into RFpeptides, a framework that generates macrocyclic peptide binders against arbitrary protein targets without any starting template. Testing 20 or fewer designed macrocycles against each of four structurally diverse proteins yielded binders with medium-to-high affinity in every case, including a sub-10 nanomolar binder generated purely from a predicted target structure. X-ray crystal structures of the bound complexes matched the computational models closely, with a Cα root-mean-square deviation under 1.5 angstroms.

This matters because macrocyclic peptides sit in a therapeutic sweet spot between small molecules and antibodies: large enough to grip flat, featureless protein surfaces that defeat small drugs, yet small enough to be synthesized chemically and, potentially, made cell-permeable. Reliable computational design collapses what used to be a slow, luck-dependent screening process into a matter of testing a handful of candidates.

A good PeptideWiki angle: a plain-language explainer on why macrocyclic peptides are the field's answer to "undruggable" targets, using RFpeptides as the worked example of how AI design now delivers nanomolar binders on the first try.

In a companion advance, Baker's team designed a complete chemically induced dimerization system entirely from scratch: a C2-symmetric, membrane-permeable macrocyclic peptide paired with a de novo protein homodimer that only assembles in the peptide's presence. Chemically induced dimerization is a workhorse of synthetic biology — used to switch signaling on, trigger cell death, or gate engineered cell therapies — but existing systems rely on a small stable of repurposed natural molecules like rapamycin.

Building both the small-molecule trigger and its protein partners de novo means researchers can design orthogonal switches that do not cross-react with anything in human biology, an important safety and specificity feature for future cell and gene therapies. It is also a proof of principle that peptide chemistry and protein design can now be co-engineered as a single system rather than bolted together after the fact.

A good PeptideWiki angle: a short piece on "designer molecular switches," framing why a peptide you can draw on a computer and a protein built to recognize it opens a cleaner path to programmable cell therapies.

A bioRxiv preprint reports a method for artificially recombining nonribosomal peptide synthetases — the modular enzymatic assembly lines that bacteria and fungi use to build many of our antibiotics, immunosuppressants, and anticancer agents. By shuffling the modules of these megaenzymes in a programmable way, the authors accelerate the evolution of entirely new bioactive peptide scaffolds rather than waiting for nature to stumble into them.

Nonribosomal peptides include clinical mainstays like vancomycin and cyclosporine, but engineering their biosynthetic machinery has long been frustrated by how tightly the modules are wired together. A recombination strategy that reliably yields functional hybrids could reopen natural-product discovery at a time when the antibiotic pipeline is dangerously thin.

A good PeptideWiki angle: an accessible primer on nonribosomal peptides — the class behind vancomycin — and why letting scientists remix the assembly line is a promising route to next-generation antibiotics.

Researchers have reported a computational framework for designing alpha-helical peptides that spontaneously insert into lipid membranes and assemble into barrel-stave pores — ring-like channels built from individual helices standing side by side. The designed pores were tuned for antimicrobial activity, punching holes in bacterial membranes, but the same controllable architecture points toward applications in nanopore sensing and targeted drug delivery.

Membrane-spanning peptide assemblies are notoriously hard to design because the energetics of burying a helix in a lipid bilayer are unforgiving. Getting de novo peptides to form defined, functional pores on demand gives the field a programmable way to make membrane-active agents, from new antibiotics to molecular sensors.

A good PeptideWiki angle: explain how a barrel-stave pore works and why designing membrane-piercing peptides from first principles is both an antibiotic strategy and a nanotech building block.

Eli Lilly's retatrutide, a triple agonist that simultaneously hits the GLP-1, GIP, and glucagon receptors, reported TRIUMPH-2 results showing up to 20.8% average weight loss and a 1.6 percentage-point reduction in HbA1c across 1,152 patients with obesity and type 2 diabetes. It is a notable data point because weight loss in people with diabetes typically lags behind that seen in patients without it, yet retatrutide's numbers remain among the strongest in the class.

The result adds to a TRIUMPH program that has now posted weight-loss figures in the high 20s in obesity-only populations, positioning retatrutide ahead of tirzepatide and semaglutide on efficacy. Lilly's regulatory filing is now expected in the first quarter of 2027, with prescription availability unlikely before late 2027 at the earliest.

A good PeptideWiki angle: update the retatrutide entry with the fresh TRIUMPH-2 diabetes data and a clear side-by-side of how the triple agonist stacks up against tirzepatide and semaglutide.

The FDA's Pharmacy Compounding Advisory Committee met on July 23 and 24 to weigh whether a slate of popular research peptides should be eligible for compounding under Section 503A. Across the two days the panel reviewed BPC-157, KPV, TB-500, and MOTS-c, followed by emideltide (DSIP), Semax, and Epitalon, and signaled support for opening a compounding pathway for several of them.

The vote matters because these peptides occupy a legal gray zone: widely sold and used off-label, but none are FDA-approved drugs. A formal compounding pathway would legitimize physician-supervised access for some of them while underscoring that "compoundable" is not the same as "approved" — a distinction PeptideWiki readers frequently ask about.

A good PeptideWiki angle: a clear explainer distinguishing FDA approval from compounding eligibility, using the six named peptides to map exactly where each now stands.

An analysis presented at CROI 2026 found that GLP-1 receptor agonists generally work well for people living with HIV, delivering the expected weight and metabolic benefits while also showing signals of improved liver, gut, and cardiovascular health — and even reduced smoking. The findings help close an evidence gap, since people with HIV are often underrepresented in the pivotal obesity and diabetes trials that built the GLP-1 evidence base.

Given the high rates of metabolic complications and lipodystrophy in long-term HIV care, confirming that these peptides are both safe and effective in this population expands their reach into a group that stands to benefit substantially.

A good PeptideWiki angle: a short post on GLP-1 agonists beyond diabetes and obesity, spotlighting the emerging HIV data as an example of the class's widening therapeutic footprint.

Researchers are taking a fresh look at Linus Pauling's long-dismissed claim that vitamin C can fight cancer, and the twist is pharmacokinetic: intravenous infusion reaches blood concentrations orders of magnitude higher than any oral dose, high enough to generate cell-damaging hydrogen peroxide selectively in tumor tissue. That mechanism was invisible in the original oral-vitamin trials that discredited Pauling decades ago.

The renewed interest is a reminder that a therapy's route of administration can determine whether it works at all, and that a discarded hypothesis can be worth revisiting when the underlying biology is finally measured correctly. Ongoing trials are testing high-dose IV vitamin C as an adjunct to standard cancer treatment rather than a replacement.

A new study reports that blocking the glycine transporter SLC6A20 restored disrupted brain signaling and improved social, communication, and repetitive behaviors in mouse models of autism. By raising glycine availability at the synapse, the intervention appears to correct an excitatory–inhibitory imbalance thought to underlie some autism-related phenotypes.

The finding is notable because it points to a specific, druggable molecular target for a condition that has few pharmacological options addressing its core features rather than its comorbidities. As with all rodent work, the leap to human relevance is large and unproven, but a clean molecular handle gives future therapeutic efforts somewhere concrete to start.

In a small study of 28 participants, a wearable patch shortened the time to reach REM sleep by 43 minutes and extended REM duration by roughly 16 minutes, all without medication or surgery. REM is the sleep stage most closely tied to memory consolidation and emotional regulation, and it is often the first casualty of aging, stress, and many sleep disorders.

A non-pharmacological way to nudge sleep architecture would be appealing precisely because sleep drugs tend to suppress the deeper and REM stages they are meant to restore. The sample is tiny and the effect will need replication, but the approach hints at a gentler class of sleep interventions.

The World Health Organization reports that an Ebola outbreak in the Democratic Republic of the Congo is outpacing containment efforts, with cases doubling in some areas faster than health workers can trace contacts. Ebola's high fatality rate and the logistical strain of reaching remote, under-resourced regions make early containment critical, and losing track of the transmission chain is the scenario responders most fear.

Vaccines and monoclonal-antibody treatments developed over the past decade have transformed Ebola from an almost uniformly fatal disease into a treatable one, but only when patients are found and reached in time. The WHO's warning is fundamentally about surveillance capacity rather than the tools themselves.