A Brain Recipe and a 70-Year Oversight · Morning Gist
Monday, November 25 stories · 16 min
A Brain Recipe and a 70-Year Oversight
Two neuroscientists ask how much of the brain genes actually specify; also, an ancient Martian rock, a long-missed world and a drug approval under fire.
Picked from 71 new articles across 10 sources · arrow keys for other days
Today in 30 seconds
30s left
Ila Fiete argues a few genetic knobs plus self-organizing rules build brain areas; smooth gene gradients versus sharp steps can test it. story 1 ↓
Zircons date Teghaza 001, a Martian meteorite found in Mali, to 4.1 billion years, and the rock shows signs of water. story 2 ↓
FDA says manipulated data underpinned Tavneos's approval; Europe pulled it, but the $220,000-a-year drug stays on US shelves while Amgen contests withdrawal. story 3 ↓
Rubin data reveal 2026 ED138, a possible dwarf planet up to 1,500 km wide, overlooked since 1954, plus new Neptune Trojans. story 4 ↓
Calling the whole brain self-organized explains nothing, says Hiesinger; retinal waves and a 15-hour fly-brain structure pass his two-part test. story 5 ↓
1/5The Transmitter· Ila Fiete· Nov 2
Neuroscientist argues the genome works as a recipe for the brain
In an essay, neuroscientist Ila Fiete argues that genes set simple rules and dials while much brain structure organizes itself. Her evidence combines limb development with her own models, whose key predictions remain untested.
Why it matters · It turns a vague debate about nature and emergence into a testable contrast: smooth gene gradients would favor self-organization, while sharp steps at brain-area boundaries would favor genetic instruction.
Know this first
Context from the editor, not the article · 27 sec
1 · Too few genes
The human genome has roughly 20,000 protein-coding genes, while the brain has on the order of 100 trillion synapses. Genes cannot list every connection individually, which raises the question of how brain structure gets specified.
2 · How embryos make patterns
Embryos turn uniform tissue into distinct parts using signaling molecules. Cells respond to local chemical levels by switching genes on or off, so a cell's position in the tissue can shape what it becomes.
3 · Grid cells
Grid cells are neurons that fire at regularly spaced locations as an animal moves, forming a hexagonal lattice across space. They sit in the entorhinal cortex, a brain region involved in navigation and memory.
2/5Universe Today· Evan Gough· Oct 30
A 4.1-Billion-Year-Old Martian Meteorite Found in Mali Shows Signs of Water
Scientists dated a meteorite found in Mali to at least 4.1 billion years old, near the start of Martian history. Its oxygen and altered minerals show it reacted with fluids before leaving Mars.
Why it matters · Mars's earliest era is the likeliest window for surface life and the hardest to study, and this rock gives laboratories a rare direct sample of its crust.
Know this first
Context from the editor, not the article · 30 sec
1 · How Mars reaches Earth
Large impacts can blast Martian rock into space, and some eventually lands here as meteorites. Deserts and ice fields yield many finds because dark stones stand out, which is why the Sahara is such a productive hunting ground.
2 · The warm, wet Mars question
Mars is now cold and dry, but valleys and water-altered minerals suggest it was once wetter. How warm, how wet and for how long remain disputed, partly because so little rock from the earliest era has been sampled.
3 · Isotopes as chemical fingerprints
Isotopes are versions of an element with different masses. Radioactive ones decay at known rates, so they work as clocks. Stable ones, such as the oxygen isotopes, shift in proportion during reactions, recording what a rock has interacted with.
3/5Ars Technica Science·
Sara Talpos, Undark Magazine
· Oct 31
FDA says a rare-disease drug's approval rested on manipulated data; removing it is slow
The FDA proposed withdrawing Tavneos, a vasculitis drug, alleging its maker manipulated trial data and citing liver injuries. Owner Amgen denies manipulation and is fighting back, so US patients face prolonged uncertainty about whether it works.
Why it matters · It tests what protects medical evidence when peer review and regulators both miss an alleged manipulation, and what regulators can do once patients depend on a drug.
Know this first
Context from the editor, not the article · 22 sec
1 · How US drug approval works
Companies usually run and analyse the large trials behind an application, then submit results to the FDA. The law requires substantial evidence of safety and effectiveness. Outside advisory committees can weigh in, but the FDA decides, so the system relies heavily on honest reporting.
2 · Blinding and adjudication
In a blinded trial, those judging outcomes don't know who received the drug. Judgment calls, such as whether a patient reached remission, are made by set rules, often by an adjudication panel. Blinding keeps hopes for a result from influencing those calls.
Image: Ars Technica Science
4/5Science News· Lisa Grossman· Oct 30
Rubin Observatory finds a possible dwarf planet hiding in plain sight since 1954
A new telescope in Chile has spotted an object beyond Neptune, possibly 600–1,500 kilometers wide and likely a dwarf planet. Astronomers then found it in images from 2016 and 1954.
Why it matters · A body this large went unnoticed for at least 70 years, which suggests how incomplete the census of the outer solar system remains and what systematic surveys can now fill in.
Know this first
Context from the editor, not the article · 27 sec
1 · The Kuiper Belt
A broad ring of icy bodies beyond Neptune, roughly 30 to 50 times Earth's distance from the sun. Its objects are faint, slow-moving leftovers from the solar system's formation, so many remain undiscovered.
2 · What makes a dwarf planet
Under the International Astronomical Union's 2006 definition, a dwarf planet orbits the sun and has enough gravity to pull itself round, but has not cleared other objects from its orbit. Pluto is the best-known example.
3 · How moving objects are found
Stars stay fixed relative to each other, while solar system objects shift between exposures. Surveys compare repeated images of the same patch of sky, then link the moving dots into a path that fits an orbit.
5/5The Transmitter· P. Robin Hiesinger· Nov 2
A neurobiologist's test for when self-organization actually explains brain wiring
In an opinion essay, P. Robin Hiesinger argues self-organization explains brain wiring only when interacting parts produce rules none has alone, through a process in time. Retinal waves qualify; calling whole brains self-organized doesn't.
Why it matters · It offers a way to separate a real explanation from a fashionable label, and raises the question of how much brain wiring genes specify directly.
Know this first
Context from the editor, not the article · 30 sec
1 · The brain wiring problem
Brains contain far more synapses than genomes have genes, so DNA cannot plausibly list each connection. Developmental biologists have often explained wiring through molecular cues, such as chemical gradients and target molecules that attract or repel growing axons.
2 · Hebbian learning
This is the principle that a synapse strengthens when the neurons on both sides of it are repeatedly active together, while poorly correlated connections weaken or disappear. It is often summarized as "cells that fire together wire together."
In an essay for The Transmitter, neuroscientist Ila Fiete argues that her field has treated the genome too much like a blueprint for the brain. She prefers a box of cake mix. A cake's crumb and aroma aren't in the box; they emerge once water and heat set chemistry going.
Genes, she writes, supply ingredients and brief directions. They don't specify the border between two visual areas or the wiring between each pair of neurons. Those arise through 1, in which simple local processes, repeated many times, build complex structure. This is an opinion piece, not a new experiment, and her key predictions remain untested.
Two ways to build a body
In the blueprint picture, Lewis Wolpert's 2 has a chemical gradient across a tissue, with genes switching on at set concentrations to tell cells what to form.
At the other extreme is the 3, named for Alan Turing: one chemical activates itself locally while another inhibits over a longer range, and stripes or spots appear spontaneously. Genes choose the molecules and amounts, but not where each spot goes.
Digits show the second approach at work. James Sharpe's group found that they form as a Turing pattern of the signaling components Wnt, BMP and SOX9. Varying Wnt or BMP changes the pattern's spacing, so the embryo grows more or fewer digits.
Each approach has costs. Gradient systems scale with body size but can be noise-sensitive and need more genes. Turing patterns are robust and need few genes, but they set only one scale. Fiete says the real question is which mechanism the brain uses where, and when.
What her models show
4, neurons in the 5 that help map space, come in discrete modules, yet the cell properties beneath them appear to vary smoothly. Fiete and colleagues showed in a model that two local interactions, one graded and one of fixed width, can create sharp module boundaries. In this mechanism, called 6, a bigger brain gets bigger modules, not more of them, unlike a Turing pattern.
A second model, built with Talia Konkle, has neurons sprout synapses according to their activity and compete for a limited construction budget, with only spontaneous waves of activity as input. It produced discrete visual areas, a hierarchy and 7, in which neighboring points in the visual field map to neighboring neurons.
The test
That model predicts only smooth gene-expression gradients at higher visual-area boundaries, not sharp steps. 8, which measures gene activity across intact tissue, can now check this.
If she's right, changing the number of cortical areas could mean turning one genetic dial. Fiete suspects this is also why brains can evolve: selection can tune a dial far more easily than it can coordinate many edits to a blueprint.
Neuroscientist Ila Fiete argues, in an essay for The Transmitter, that her field has leaned too far toward treating the genome as a blueprint for the brain. Her alternative image is a box of cake mix. A cake's crumb and aroma are nowhere in the box. They appear once water and heat set off processes such as proteins tangling, starches gelling and gas bubbles expanding.
Genes, she writes, similarly supply ingredients and a short list of directions: what to combine, how much and when. They don't specify the border between the secondary visual cortex and its neighbor, the wiring between each pair of neurons, or the exact size of a person's cortical sheet. Those emerge through 1, in which simple local processes, repeated many times, build complex structure.
This is a position piece, not a new experiment. It draws on others' findings and on models from Fiete's own lab, and it reports no quantitative fits for those models. Her most distinctive predictions are still waiting to be tested.
A pendulum that swung
The idea is old. In the mid-20th century, Alan Turing and John von Neumann showed that simple local rules, applied repeatedly, can generate rich structure. Then the genetic revolution of the 1980s gave embryologists concrete molecular information and tools, along with a strong intuition that genes specify function. Fiete wants the pendulum to swing back, without displacing genetics. In her modified view, genes set the rules and turn the knobs, and self-organization does much of the rest.
Two ends of a spectrum
Two classic ideas represent the blueprint end. In Roger Sperry's 2, axons, the long output fibers of neurons, find their targets by matching molecular labels. In Lewis Wolpert's 3, a molecule's concentration falls off across a tissue, and downstream genes switch on at particular thresholds, telling cells what to form at each location.
The textbook case is the fly embryo. A gradient of a protein called bicoid switches on the 4, which activate stripes of pair-rule genes, leading to body segments.
At the other end is the 5. Two interacting chemicals, one that activates itself locally and one that inhibits over a longer range, spontaneously break the symmetry of a uniform tissue and produce stripes or spots. Nothing genetic says where each spot goes. The genome acts as a cookbook, choosing which molecules to mix and in what concentrations.
Digits are Fiete's main example from body development. James Sharpe's group showed that digits form as a Turing pattern of three signaling components, Wnt, BMP and SOX9, in the limb bud. Varying Wnt or BMP signaling changes the pattern's wavelength, so the embryo grows more or fewer digits. No French-flag-style set of genes marks each digit's boundary. Wnt and BMP act as dials on the spacing of a self-organized pattern.
Trade-offs, and a better question
Each approach has costs. Chemoaffinity and French flag systems scale with body size, but they can be sensitive to noise and need more genes. Turing patterns are robust and need few genes, but they specify only a single scale. So Fiete says the interesting question is which mechanism the brain uses where, and at what stage of development.
Even the fly's blueprint looks blurry. Thomas Gregor and colleagues report that several gap genes acting together, through interactions, specify a cell's position better than a single gradient without interactions. Physical forces contribute too: Boris Shraiman, Kenneth Irvine and others have shown that cells sense stretch and compression at tissue boundaries, and that these forces influence growth and cell fate. 6, lab-grown clumps of brain-like tissue, offer a vivid demonstration. Human stem cells given a few growth factors form layered, cortex-like tissue, with cell types arising in roughly the right order.
Sharp modules from smooth gradients
Fiete's own example starts with a puzzle about grid cells, neurons in the 7 that help map space. They come in a small number of discrete modules, each with its own spatial scale. Yet underlying cell properties, such as ion channel densities, appear to vary smoothly along the dorsoventral axis, the top-to-bottom direction of the region. How can a smooth gradient produce sharp jumps?
Fiete, Mikail Khona and Sarthak Chandra found that two local interactions, one smoothly graded and one of fixed width, could be enough to produce modules with abrupt boundaries. They call the mechanism 8.
Its scaling is what separates it from Turing patterning. As an illustration of the logic, take a tissue that doubles in size. A Turing pattern keeps its spacing, because the interactions set that scale, so it gets about twice as many repeats. Under peak selection, the number of modules stays the same and each module grows with the brain. Future 9, the mapping of neural wiring at the level of individual synapses, could test the model in the entorhinal cortex.
Visual areas without a map
A second model, developed with Talia Konkle, moves up a level. Neurons in a simulated sheet of cortex sprout synapses according to their activity, and the synapses on each neuron compete for a limited construction budget. The only input is spontaneous waves of activity. This produced discrete visual areas, a processing hierarchy and 10, in which neighboring points in the visual field map to neighboring neurons, as in macaque and human brains.
The model makes a sharp prediction: no steps or switches in gene expression at the boundaries of higher visual areas, only smooth gradients. 11, which measures gene activity across intact tissue, can now check this.
Why knobs would matter
If Fiete is right, a handful of growth rules and knobs links gene networks to circuits and behavior. Changing the size or number of cortical areas would mean turning one dial, without rewriting a wiring diagram. She also suspects this explains why brains can evolve. A blueprint would need many coordinated edits for a meaningful change, she argues, while a small change to a recipe's dial yields a coherent variation across the organism. That evolutionary claim is her inference, which she flags as a suspicion.
Her colleague P. Robin Hiesinger thinks even the recipe gives genes too much credit, since the order and timing of developmental steps can themselves emerge, as in the assembly of the sorting plate that guides fly eye neurons into the brain.
What remains open is how far this reaches in real brains. Fiete calls for theorists who ask which minimal rules could produce a given structure, experimentalists who measure hypothesized knobs such as gradient slopes and interaction widths, and analysts who look for smooth gradients versus steps and switches in the data.
Remember
Fiete's framing isn't genes versus emergence: genes pick the rules and dial settings that self-organization then plays out.
A Turing pattern and peak selection respond differently to growth: one adds repeats, the other enlarges a fixed set of modules.
Even the fly embryo, the textbook genetic blueprint, positions cells better when gap genes interact than with a single gradient.
The largest of Teg 001's eight pieces, and a cut section of a smaller chunk showing its interior.1 / 3
A meteorite found in Mali in 2022 has turned out to be at least 4.1 billion years old, formed when Mars itself was only a few hundred million years old. Its chemistry also shows it reacted with fluids before leaving the planet. The peer-reviewed study of Teghaza 001 (Teg 001) appears in Science, led by Yang Liu of NASA's Jet Propulsion Laboratory.
Why this rock is rare
Ancient rock dominates the Martian surface, yet most Martian meteorites are only a few hundred million years old. An impact must blast a rock into space intact, which favors strong, young volcanic rock over old crust that has been shattered by impacts or is made of weak sediment and clay.
The best-known exception, 1, found in 2011, has parts about 4.43 billion years old. But it is a 2, a jumble of many rock types welded together. Teg 001 is one 3 rock (solidified from magma), recovered in eight pieces totalling almost 800 grams. Carl Agee of the University of New Mexico, a leading figure in the work, compares it to chipping a piece from a 4-billion-year-old outcrop.
How the team knows
Age. The team made ten measurements of uranium and lead isotopes in six crystals of 4, a durable mineral. Uranium decays to lead at a known rate, so the ratio works as a clock, a method called 5. The grains were damaged by impact, and the paper reports the age as a minimum.
Water. The rock is rich in 6, a heavy oxygen isotope that water-rock reactions tend to concentrate in minerals. The authors suggest its magma either absorbed crust that had interacted with surface fluids or reacted with such fluids directly. Crystals of 7, a common volcanic mineral, were also heavily altered by fluid.
Manganese. Agee found patches high in 8, also seen by the Curiosity rover at Gale Crater. It may point to oxygen-rich groundwater, though that remains debated.
What it does and doesn't show
The age places the rock at or near the 9, Mars's earliest era, when it may have had surface water, a denser atmosphere and perhaps a magnetic field. Agee argues the rock "probably came from an environment on Mars that was more Earth-like," and that abundant surface life, if it ever existed, most likely arose then. The rock itself offers no evidence of life.
The paper's own claim is narrower: the rock reacted with oxygen-18-rich fluids before it was ejected. The article gives no date for that alteration, and fluids could reflect local volcanic heating rather than a warm climate. Agee also calls it the oldest Martian igneous rock "by a huge margin," though the article omits 10, another igneous Martian meteorite usually dated to about 4.09 billion years.
A meteorite found in Mali in 2022 is among the oldest pieces of Mars in any laboratory, and it carries chemical signs of contact with water. In a peer-reviewed paper in Science, a team led by Yang Liu of NASA's Jet Propulsion Laboratory reports that Teghaza 001, or Teg 001, is an 1 rock (one that solidified from magma) that formed at least 4.1 billion years ago. Mars itself is roughly 4.5 billion years old. The meteorite was recovered in eight pieces totalling almost 800 grams.
Why old Martian rock is so hard to get
Ancient rock dominates the surface of Mars, yet most Martian meteorites are only a few hundred million years old. The authors say such samples are "not representative" of the surface.
The mismatch is a selection effect. A meteorite reaches Earth only if an impact launches it off Mars and it survives the blast. Young volcanic basalts are strong enough. Older crust has been fractured and brecciated, meaning shattered and re-cemented, by repeated impacts, and some of it is sedimentary or clay-bearing. Those weaker rocks are much less likely to make the trip.
The previous benchmark was 2, formally Northwest Africa 7034, found in the Western Sahara in 2011. Parts of it are about 4.43 billion years old, older than Teg 001's minimum age. But Black Beauty is a 3, a jumble of many rock types welded together, so its oldest fragments have lost their surroundings. Teg 001 is one slowly cooled rock type. Carl Agee, director of the University of New Mexico's Institute of Meteoritics and a leading figure in the study, likens it to going back in time and "chipping off a hunk of an outcrop."
How the age was measured
The date comes from 4, a tiny, durable crystal that can survive geological processes that destroy other minerals. The team made ten measurements of uranium and lead isotopes in six zircon grains, a method known as 5.
In general, the method works because zircon takes up uranium as it forms but largely rejects lead, so most of the lead inside comes from radioactive decay. Uranium-238 decays to lead-206 with a half-life of about 4.5 billion years, while uranium-235 decays to lead-207 with a half-life of about 700 million years. Two clocks running at different speeds allow a cross-check. As an illustration, a zircon 4.1 billion years old would have lost about 47 percent of its original uranium-238.
The team used both uranium-to-lead ratios and lead-to-lead ratios. Their grains were "shocked," or damaged by impact, and the paper's title gives the age as "≳4.1 billion years," meaning a floor. The article doesn't explain that choice, but shock can generally let lead escape, which makes crystals look younger than they are.
Three clues to fluids
Oxygen isotopes. Geochemists report the share of heavy 6 as 7, the departure from a reference standard in parts per thousand. Water-rock reactions at low temperatures tend to concentrate oxygen-18 in the resulting minerals, so altered crust ends up isotopically heavy. Teg 001's value is high, though the article gives no figure. The authors offer two explanations: its magma absorbed crust that had interacted with surface fluids, or the magma reacted directly with oxygen-18-rich fluids.
Altered minerals. Pigeonite is a pyroxene, a common silicate in volcanic rock. "The Mg-rich interiors of 8 have been heavily altered by reactions with a fluid phase," the authors write, meaning the magnesium-rich cores of these crystals were chemically changed.
Manganese oxide. Agee found microscopic domains high in 9, which he describes as a compound rovers look for as an indicator of environments conducive to life. The Curiosity rover found fractures rich in it at Gale Crater, a result that suggests, but doesn't prove, free oxygen in the atmosphere and oxygen-rich groundwater. That reading is still debated.
The paper's own summary is measured: altered minerals and oxygen isotope measurements indicate that Teg 001 "experienced reactions with fluids enriched in 18O, before ejection into space by an impact."
What it would mean
Teg 001's age puts it at or near the 10, the earliest chapter of Martian history, largely erased by later geology and weathering. Some research suggests Mars may have had global oceans then.
Agee goes further than the paper's summary. He says the sample "probably came from an environment on Mars that was more Earth-like," with surface water and volcanoes driving 11, where heated water circulates through rock. Maybe, he adds, Mars then still had a magnetic field shielding it from radiation, plus a denser atmosphere. If life was ever abundant on the surface, he argues, that is when it most likely happened. Nothing reported here is evidence of life itself; this is an argument about where to look.
The rock's source may also matter. It likely came from a place no rover has visited, possibly the southern highlands, an ancient region whose composition, measured by remote sensing, resembles Teg 001.
What remains open
Several gaps separate the measurements from the "warm and wet" headline.
Timing. "Before ejection" is a wide window, and the article gives no date for the alteration or for the launch from Mars. Oxygen absorbed by the magma would be about as old as the rock, but the altered pigeonite could have changed later, possibly much later.
Remember
Martian meteorites skew young because impacts launch strong volcanic rock more readily than old, fractured crust.
Teg 001's value comes from being one intact ancient rock type, unlike the older but jumbled Black Beauty.
Fluid alteration is measured; a warm, Earth-like Mars with possible surface life is the researchers' broader interpretation.
In April, the US Food and Drug Administration (FDA) proposed withdrawing Tavneos, a drug for a rare autoimmune disease, saying its approval rested on manipulated trial data. The European Union has since removed it, and the UK says it will. In the US it is still sold, at about $220,000 a year, while its owner, Amgen, fights the proposal.
A need and a weak case
Tavneos treats 1 (AAV), rare diseases in which the immune system attacks blood vessels, often damaging the kidneys and lungs. Patients need high-dose steroids, which work fast but can cause severe weight gain, mood swings and insomnia, sometimes for years. Tavneos was meant to reduce that burden.
The evidence was disputed from the start. A trial by ChemoCentryx, a small California biotech, showed slightly higher remission rates at one year. But critics noted that many comparison patients got one round of 2, a slower-acting drug, where real-world care would have used two. An independent 3 narrowly backed approval anyway, and the FDA agreed in October 2021. Amgen bought ChemoCentryx for nearly $4 billion a year later.
What a lawsuit uncovered
A shareholder suit produced a 2025 report alleging that the company's own analysis originally found no remission advantage. Employees then 4 the data, checking who got which treatment, and chose a handful of patients for reassessment, a step the FDA calls 5. The FDA says this broke the trial's protocol and alleges the patients were cherry-picked. Neither the first result nor the reassessment was disclosed to the agency.
This matters because someone who knows the assignments can pick which borderline cases to revisit, and that choice alone can create a difference between groups. The New England Journal of Medicine retracted the trial paper this summer at its academic authors' request.
The FDA also cites liver injury. Its acting drug chief, Tracy Beth Høeg, says Tavneos has been linked to eight deaths and dozens of hospitalizations.
The defense, and the slow process
Amgen, which says it didn't know about the reassessment, calls the nondisclosure "unfortunate" but denies manipulation. It says no clinical data changed and that a small number of errors were corrected. An independent re-analysis it cites found Tavneos still fell short of higher remission rates, though patients fared no worse. Routine-care data suggest lower steroid use, and one independent study found low liver-injury rates.
Skeptics note that several of Amgen's studies remain unpublished, and that it is behind on follow-up studies the FDA requested.
Withdrawal itself is a stepwise legal process: notice, the offer of a hearing if the facts are genuinely disputed, then a decision Amgen could challenge in court. Until then, patients and doctors are left without a clear answer on whether the drug works.
In April, the US Food and Drug Administration (FDA) proposed withdrawing its approval of Tavneos, a drug for a rare autoimmune disease. The agency argued that the drug now lacks "substantial evidence" that it works, because the trial behind its approval appears to have been manipulated, and it cited reports of serious, sometimes fatal, liver injury.
Events moved quickly after that. This summer the New England Journal of Medicine retracted the trial paper, and the European Union took the drug off the market. The UK has said it will follow. In the US, though, Tavneos remains available at about $220,000 a year, and its owner, Amgen, is contesting the proposal. Reporting for Undark, Sara Talpos traces how this happened and why it is hard to undo.
The disease and the unmet need
Tavneos treats 1 (AAV), a group of rare diseases in which the immune system attacks the body's own blood vessels. It can rapidly cause organ failure, particularly in the kidneys and lungs. AAV was usually fatal when first identified, but drugs developed over four decades now let many patients live a relatively normal lifespan.
Standard care starts with high-dose steroids to quell inflammation fast, plus slower-acting drugs aimed at the underlying disease. Steroids work but can be punishing, and patients may need them for months to years. Chloe Cusano, diagnosed at 20, recalls gaining 40 pounds in about six weeks, along with violent mood swings, insomnia and nightmares. Tavneos was supposed to replace or reduce steroid use.
A contested approval
The evidence looked uncertain from the start. The trial, run by a small California biotech called ChemoCentryx, showed slightly higher remission rates at one year. In May 2021, however, an FDA briefing document said the study's design made those results hard to interpret.
David Pisetsky, a Duke rheumatologist on the FDA's independent 2, voted no. One reason was that many patients in the comparison group got one round of 3, a slower-acting drug, when in real-world care they would have received two. A weaker comparison makes a new drug look better. Mike Putman of the Medical College of Wisconsin later made a similar argument in an editorial, while conceding that the standard of care at the time was somewhat debatable.
Walter Kraft, a clinical pharmacologist at Thomas Jefferson University, voted yes, judging that limited evidence met the need and that another trial could delay the drug by years. His vote, he now says, was "based upon an assumption that the sponsor was following the rules of the road."
The committee narrowly backed the drug, and the FDA approved it on October 7, 2021. A year later Amgen bought ChemoCentryx for nearly $4 billion. Tavneos sales reached nearly half a billion dollars in fiscal 2025.
What the lawsuit uncovered
The evidence of a problem came almost by accident. ChemoCentryx's stock fell after the skeptical briefing document appeared, and shareholders sued for securities fraud. A report filed in that case four years later alleged that the company's own analysis had originally found no better one-year remission with Tavneos than with usual treatment.
According to the report, employees then 4 the data, meaning they looked at which participants had received which treatment, and selected a handful of patients for reassessment. The FDA calls this 5 and says it violated the study protocol. It also alleges that the patients were cherry-picked to produce the result needed for approval. Neither the original result nor the reassessment was disclosed to the agency.
An illustrative example, not using the trial's figures, shows why a handful can matter. Suppose each arm has 100 patients and 70 in each reach remission. Moving three drug-arm patients to "remission" and three comparison patients out of it yields 73 versus 67. Each change might look defensible on its own. But a reviewer who knows the treatment assignments can choose which borderline cases to revisit, and that choice alone can create a difference.
Amgen's defense
Amgen says it was also unaware of the reassessment. It concedes that the work doesn't "align with the approach Amgen would have taken" and calls the nondisclosure "unfortunate," but denies that it amounts to manipulation. In a 104-page submission, it says no underlying clinical data changed, that the reassessment corrected a small number of errors under a prespecified procedure, and that ChemoCentryx had no legal duty to disclose it.
Tracy Beth Høeg, who became acting director of the FDA's drug center in December 2025, takes the opposite view with unusual bluntness. "Essentially ChemoCentryx swindled billions of dollars," she wrote to Undark, adding that the drug has been linked to eight deaths and dozens of hospitalizations. "This can be a deadly drug, and we don't know if it works."
Why withdrawal takes time
The FDA can withdraw approval, but only in steps, explains Patricia Zettler, an Ohio State law professor. The agency must give notice and offer the manufacturer a chance at a public hearing. If the manufacturer shows a genuine dispute over the facts, the hearing may be held, and only then can the FDA withdraw approval. Amgen could then sue.
What evidence remains
An independent re-analysis by the Duke Clinical Research Institute, cited by Amgen, found Tavneos still didn't meet the FDA's bar of higher one-year remission rates, though patients on it fared no worse. Amgen also points to 6, collected during routine care, suggesting reduced steroid use. Eric Roberts, a University of California, San Francisco epidemiologist, says his independent 7 found low rates of drug-induced liver injury, which he calls "incredibly reassuring," especially with careful monitoring.
Skeptics note the limits of this evidence. Joshua Skydel, a Yale researcher and rheumatology fellow, says several of Amgen's studies are unpublished and he wants to see them peer-reviewed. Amgen has also fallen behind on 8, the follow-up work the FDA requested at approval. Høeg suspects real-world data won't meet the agency's evidentiary bar.
What stays open
The article doesn't settle whether Tavneos works, and it can't. The one randomized trial is compromised, and the replacement evidence is contested. Amgen plans to present new data in November. A hearing on a proposed $69 million shareholder settlement was scheduled for October 29, and the article doesn't report the outcome.
The wider worry is structural. Economists note that many new drugs now start at small companies, and David Ridley of Duke says those firms can have everything riding on one product, so "in rare cases, they might cut corners." Putman argues that peer review and regulatory approval both failed to catch this. The FDA says that leaving the drug on the market would give other companies an incentive to deceive.
Remember
The alleged problem surfaced through a securities lawsuit, not through peer review or FDA review of the trial.
No data values needed changing: choosing which patients to reassess after unblinding can itself shift a result.
Even the re-analysis Amgen cites shows no remission advantage, shifting its case to steroid reduction and real-world evidence.
1954 Palomar plate, first of a pair: ED138 is the faint dot circled in red at center.1 / 2
The 1, a telescope in Chile built to map the southern sky every couple of nights, has turned up a possible new 2 less than five months into its 10-year survey. The object, 2026 ED138, was reported this week at a planetary science meeting. It has not yet been confirmed as a dwarf planet.
Found in data already collected
Rubin recorded the object 30 times on 20 nights between March and July, but nobody noticed until October 13, when astrophysicist Matt Holman ran a search for moving dots of light.
ED138 lies about 47 times as far from Earth as Earth is from the sun, in the 3, the ring of icy bodies beyond Neptune. Its brightness suggests a width of 600 to 1,500 kilometers. The range is wide because brightness depends on surface reflectivity as well as size. An object in that range is probably massive enough for gravity to have made it round, the trait that would make it a dwarf planet. Amateur astronomer Sam Deen calls it the biggest thing found in the solar system in 19 years.
How astronomers checked
Holman was thrilled but doubtful, so the team looked for the object in older images. They found it in 2016 data from the 4, an instrument on a neighboring Chilean mountaintop. Deen and other amateurs then traced it to 5 taken at California's Palomar Observatory in 1954.
Those old images pin down the orbit. ED138 takes 280 years to circle the sun, so a few months of data cover only a sliver of its path, while 72 years cover about a quarter. It stayed hidden because it moves slowly and was crossing a dense field of Milky Way stars.
The precise orbit let Deen plan to watch an 6 on Halloween, when ED138 would pass in front of a distant star. Timing how long the star disappears can reveal the object's size and shape, and any moons or rings. Such predictions usually take years of orbit tracking. The article appeared before the event, so the result isn't known.
New Trojans behind Neptune
A second team, publishing in Astrophysical Journal Letters, searched Rubin's early test data for 7, icy bodies sharing Neptune's orbit about 60 degrees ahead of or behind it. Astronomers knew of 30 ahead but only two behind, likely because the trailing group has been crossing the Milky Way for 15 years. The team found two more trailing objects plus one candidate. Meg Schwamb of Queens University Belfast expects hundreds more to emerge.
Astronomer Sam Lawler of the University of Regina warns that the growing number of bright satellites will make such finds harder, citing 100 new launches a week.
Less than five months into its decade-long survey, the 1, a telescope in Chile built to map the southern sky every couple of nights, has produced a possible new 2. Astronomers described the object, called 2026 ED138, this week at a meeting of the American Astronomical Society's Division of Planetary Sciences in Spokane, Washington. A second team reported new 3, icy bodies that share Neptune's orbit, at the meeting and in the peer-reviewed Astrophysical Journal Letters. The article describes the ED138 result only as a meeting report, so treat it as early.
A find already sitting in the data
Rubin took its first images in June 2025 and formally began its survey in June 2026. It recorded ED138 30 times on 20 nights between March 4 and July 11, but nobody knew the object was there. On October 13, astrophysicist Matt Holman of the Smithsonian Astronomical Observatory ran a search for moving dots of light and found it.
ED138 appears to be about 47 times as far from Earth as Earth is from the sun, in the 4, the ring of icy bodies beyond Neptune. From its brightness, astronomers estimate a width between 600 and 1,500 kilometers, which could put it in the company of the dwarf planets Ceres and Makemake. An object that size is probably massive enough for gravity to have pulled it round, the property that would make it a dwarf planet. That has not been shown directly, which is why it remains a "potential" dwarf planet.
The wide size range comes from a basic limitation. Brightness measures reflected sunlight, which depends on both an object's area and how reflective its surface is. A small, icy, shiny body can look the same as a large, dark one. By simple geometry, the top of the range is 2.5 times as wide as the bottom, so it needs about six times less reflectivity to appear equally bright.
Amateur astronomer Sam Deen was emphatic about the scale: "No matter what you call it, it's still the biggest thing we've found in the entire solar system in 19 years."
Checking it against old images
Holman says he was thrilled but doubtful, and that a claim of this size has to be verified. The check was to look backward. If the object is real and its predicted orbit is right, it should appear at the calculated positions in older images.
It did. Holman's team found it in 2016 data from the 5, an instrument on an adjacent Chilean mountaintop. Deen then traced it to 6 from a 1954 sky survey at Palomar Observatory in California, and other amateurs independently found the same sightings. Two Palomar exposures, taken about half an hour apart, happened to point at the right spot. "If you know it's supposed to be there, it's not particularly faint," Deen says.
This long record matters for the orbit. ED138 circles the sun once every 280 years. By our arithmetic, Rubin's four months of observations cover only about 0.1% of a full orbit, whereas the stretch back to 1954 covers about a quarter of one. A short arc fits many possible orbits, while a long one rules most of them out.
The same slowness helps explain how the object stayed hidden. It moves only about 1.3 degrees per year along its orbit, again by our calculation, and it was crossing a dense field of Milky Way stars, where it blended in.
The Halloween 7
The precise orbit makes an unusually early follow-up possible. Deen hoped to watch ED138 pass in front of a distant star on Halloween, an event called an occultation. Usually it takes years to know an orbit well enough to predict one, he says.
The method is direct. As the object blocks the star, its shadow sweeps across Earth, and an observer times how long the star disappears. As an illustrative example, a shadow moving at 20 kilometers per second that takes 40 seconds to pass implies a line 800 kilometers long across the object. Observers at different locations measure different slices, which together outline its size and shape. Unlike the brightness estimate, this does not depend on reflectivity. Occultations can also reveal moons or rings, if they exist. The article was published before Halloween, so it doesn't say whether the attempt worked.
Evening up Neptune's Trojans
The Milky Way also hid the second discovery. Trojans gather in a planet's orbit, about 60 degrees ahead of or behind it. Jupiter has thousands. Astronomers think Neptune has many more than the handful known, and consider them original building blocks of the solar system, preserving a record of how the planets formed and evolved.
The count has been lopsided, with 30 known in the cloud ahead of Neptune and just two in the cloud behind. "We would expect these [groups of Trojans] to be symmetric in numbers," said planetary scientist Meg Schwamb of Queens University Belfast. The likely reason is an accident of geometry: for the last 15 years, the trailing cloud has appeared in front of the Milky Way as seen from Earth. Schwamb compared the search to "finding a needle in a sequined dress."
Her team searched data from Rubin's 8, when engineers were still testing the telescope, and found two more trailing Trojans plus one candidate. That doubles the confirmed trailing count, though it remains far short of 30. The new objects seem similar in composition to the leading ones, which matches theories of solar system formation, though the article doesn't give details of those measurements. "These objects photobombed the commissioning survey," Schwamb said, predicting that hundreds more are likely to be found.
What could limit the next finds
Astronomer Sam Lawler of the University of Regina cautions that satellites crossing Rubin's field of view will make these discoveries harder. "There are one hundred new satellites launching every week, and they keep getting bigger and brighter," she says. The article doesn't quantify how many observations would be lost.
Still, she expects more. ED138 turned up quickly, and Lawler speculates that something as large as Mars could be out there, which would reopen the debate over what counts as a planet. That is a possibility rather than a prediction backed by any data here. For now, the open questions are concrete ones: how large ED138 really is, whether it is round, and whether it has moons or rings.
Remember
Size estimates from brightness are weak: without knowing surface reflectivity, ED138's possible width spans a factor of 2.5.
Archival images can be as useful as new ones: 1954 plates extended ED138's orbital record by decades, enabling an early occultation forecast.
Neptune's lopsided Trojan count, 30 ahead and two behind, likely reflects a viewing bias from the Milky Way.
Ordered patterns can arise in physical systems without any template or central controller, driven by a flow of energy and local interactions. Heat a thin layer of liquid from below and it can form regular rolling convection cells.
Image: The Transmitter
In an essay for The Transmitter, developmental neurobiologist P. Robin Hiesinger asks when "self-organization" really explains how a brain wires itself. This is an argument, not a new finding, and it reports no new data.
His target is loose usage. 1, lab-grown tissues that develop from reprogrammed stem cells into patterned brain structures, are often called beautiful examples of self-organization. Hiesinger says that saying a whole brain self-organizes is plausible but explains nothing, like saying development rests on molecular interactions. A real mechanism must show how a system increases its internal order, moving from random connections to specific ones.
The model case
His example is the retinal wave. In the developing vertebrate eye, before light-sensing cells mature, 2 (the neurons that connect eye to brain) spontaneously generate waves of electrical activity that sweep in random directions. Theorists predicted this in the 1970s, and it was discovered in the 1990s.
Because a wave reaches adjacent cells almost together, neighbors fire together more often than distant cells. 3, the rule that connections strengthen when cells are active together, then strengthens neighboring synapses in the brain and prunes those farther apart. Imprecise wiring becomes an ordered pattern.
No individual cell knows about the wave. Each only excites its neighbors and reacts to them. This resembles what physical chemists call 4, ordered patterns such as whirlpools that persist by dissipating energy. Another comparison is 5, the regular patterns that form in a heated liquid. "There is no such thing as a wave at a single moment," Hiesinger writes.
Two criteria
He proposes that self-organization explains something only when:
New rules emerge from interactions among components, and the components alone do not reveal them.
The mechanism is a process in time, so no static picture can capture it.
He argues this is its own category of explanation. Molecular neuroscientists tend to look for specific molecules, and computational neuroscientists for algorithms. Self-organization could be common ground.
A structure that exists for hours
His second example comes from the fruit fly 6. Photoreceptor axons rewire into a specific pattern within about 15 hours. In his account, they follow no particular chemical gradient and seek no particular target. Instead, thousands of 7, the moving tips of growing axons, assemble into a temporary sheet that exists neither before nor after. Without it, wiring fails.
His analogy is fans driving to a theater that gets built only because they go looking for it.
What remains open
Hiesinger urges developmental biologists to look beyond gradients and target molecules and work more with physicists and theorists. The essay doesn't say how to test experimentally whether a wiring step meets his criteria, or how much of brain wiring works this way. Those are the questions his proposal will have to answer.
P. Robin Hiesinger, a developmental neurobiologist, has published an essay in The Transmitter asking when "self-organization" actually explains how a brain gets wired. This is an argument about what counts as an explanation, not a new experimental result. It reports no new data and cites its examples only in passing. It is paired with a companion essay by the theoretical neuroscientist Ila Fiete.
His complaint is about a word doing too little work. 1, lab-grown tissues that develop from induced pluripotent stem cells (adult cells reprogrammed to an embryonic-like state) into patterned brain structures, are often praised as beautiful examples of self-organization. Hiesinger asks where the insight is. Saying the entire brain self-organizes is plausible but explains nothing, he writes, much as it is true but unhelpful to say that all of brain development rests on molecular interactions.
The classic case: retinal waves
His model example comes from the developing vertebrate retina. Before photoreceptors, the light-sensing cells, have matured, waves of electrical activity arise spontaneously and sweep across the tissue in random directions. The waves are generated by 2, the neurons whose axons carry signals from the eye to the brain. Theorists predicted the phenomenon in the 1970s with surprising accuracy, and experiments found it in the 1990s.
The waves solve a sorting problem through 3, the rule that connections strengthen when the cells involved are active together. A wave ensures that neighboring neurons fire together more often than distant ones.
Here is an illustrative sketch. Take three ganglion cells: A and B sit side by side, while C is far away. A wave passing through A reaches B almost at once, but reaches C much later, if at all. Over many waves, the activity of A and B becomes correlated. In the brain, the synapses of neighboring cells are selectively strengthened, while those of cells farther apart are pruned. Initially imprecise connections sharpen into an ordered pattern.
Is that "genetically encoded"?
Forced into a binary of nature versus nurture, the waves count as pure nature. They appear before any sensory input and come solely from the retina's internal state and dynamics. Yet they aren't written in the genome in any direct sense. Whether the process counts as genetically encoded, Hiesinger says, depends on what "encoding" means. The waves emerge from how cells interact.
Physical chemistry has a long-standing category for this: 4, ordered patterns such as whirlpools that emerge and persist by dissipating energy. Hiesinger highlights two properties. They arise through a process that is irreversible in time. And their components create a new mechanism through their interactions, one that studying the components individually would not reveal.
Retinal waves fit. Each ganglion cell simply excites its neighbors and reacts to their activity, oblivious to the wave it belongs to. Direction and speed are properties of the interactions unfolding over time. The comparison Hiesinger draws is with 5, the organized patterns that form in a heated liquid of suitable viscosity. Neither rolls nor waves can be described by a snapshot. "There is no such thing as a wave at a single moment," he writes.
What makes it an explanation
Hiesinger's central demand is that an explanatory mechanism of development must show how a system can increase its internal order. Convection rolls are a more ordered, lower-entropy state than a uniform liquid, and a specifically wired network is more ordered than a randomly connected one. Retinal waves show how that increase comes about. The blanket statement that a whole brain or organoid "self-organizes" does not.
He then distills two criteria:
Emergent rules. New, mechanistically relevant rules arise from the interactions of a system's components, and the individual components do not reveal those rules.
A process in time. No static state or depiction of the system explains the mechanism.
Part of the resistance, he suggests, is cultural. Molecular neuroscientists find explanatory power in specific molecules and their measured properties. Computational neuroscientists find it in the algorithm, whichever cells or molecules happen to run it. The historian and philosopher of science Evelyn Fox Keller called these different "6," meaning communities with different standards for what counts as knowing something. "Your mechanism may not be my mechanism," as Hiesinger puts it. He argues that self-organization is its own category of explanation, one that theorists and experimentalists can share, and notes it is already well established for pattern formation in developmental biology.
A fleeting structure in the fly brain
To show that the idea reaches beyond retinal waves, he turns to 7, the fruit fly. During brain development, photoreceptor axons rewire into a specific connectivity pattern within about 15 hours. On Hiesinger's account, standard attractive or repulsive guidance doesn't describe this: the axons aren't seeking a particular downstream target or following a particular chemical gradient.
Instead, thousands of 8, the motile tips of growing axons, spontaneously assemble into a temporary sheet, a structure that exists neither before nor after the transformation. Without it, the brain does not wire up correctly. The sheet moves toward its new connectivity state with no target beyond its own searching dynamics.
His analogy is concert-goers driving to a venue that hasn't been built, where the theater comes into existence only if the fans go looking for it. The guidance mechanism is created by the interacting axons, and the destination exists only as a result of the process.
What the essay leaves open
The essay's call to action is modest: developmental biologists should look beyond gradients and target molecules, consider how structures self-assemble, and spend more time with physicists and theorists. Fiete asks for theorists who treat development as seriously as learning and who ask what other minimal rules could produce a given structure. Hiesinger adds that biologists must treat theory as seriously as genes and molecules.
Several questions remain unanswered. The essay doesn't say how a researcher would demonstrate experimentally that a given wiring step meets the two criteria, or how the fly observations rule out molecular guidance. It also doesn't say how much of brain wiring works this way. Those two examples are offered as illustrations rather than an estimate. The criteria are a proposed standard, and the essay's own evidence for their usefulness is those few cases and the established record of pattern-formation models.
Remember
A label explains nothing unless it shows how order increases; "self-organized" and "molecular" can both be true yet uninformative.
Spontaneous activity complicates nature versus nurture: retinal waves precede sensory input, yet are not directly written in the genome.
Scientists can disagree about mechanisms because their fields define explanation differently: molecules for some, algorithms for others.
Climate. Fluids reacting with rock don't establish a warm climate or oceans. Agee himself mentions hydrothermal systems, and volcanic heat could drive those locally. In my reading, the rock supports water near the crust more directly than it supports a warm planet.
The age claim. Agee calls Teg 001 the oldest Martian igneous rock "by a huge margin." The article doesn't mention 12, a well-known Martian igneous meteorite usually dated to about 4.09 billion years, so that phrase deserves caution, at least against a 4.1-billion-year minimum.
Origin. A compositional match from orbit suggests a source region but doesn't pinpoint one.
Sample size. As Agee notes, this is one meteorite. It's a single data point from a poorly sampled era, valuable because there are so few others.
