Science, amplified.

Rare glioblastoma variants are usually treated as ordinary glioblastoma because prospective variant-specific cohorts are scarce. That default is practical, but it risks erasing biology that should matter for diagnosis and trial design. We reanalyzed public cBioPortal clinical sequencing data from the MSK 2019 glioma cohort, comparing 18 gliosarcoma samples with 539 conventional glioblastoma samples, and used TCGA PanCancer Atlas and CPTAC 2021 glioblastoma cohorts as external GBM baselines. Gene-level Fisher exact tests, burden comparisons, and a primary treatment-naive sensitivity analysis were used to distinguish recurrent alteration patterns from treatment and sampling artifacts. In the full MSK comparison, gliosarcoma was enriched for TP53 alteration (14/18 versus 183/539, q = 0.0067) and RB1 alteration (8/18 versus 78/539, q = 0.0258), and depleted for EGFR alteration (1/18 versus 219/539, q = 0.0258). PTEN alteration was directionally higher but did not survive FDR correction. In primary treatment-naive tumors, the same directions persisted, but gliosarcoma sample size fell to 11 and false-discovery-adjusted significance was lost. Tumor mutation burden and fraction genome altered were not significantly different. These data support gliosarcoma as a pathway-distinct glioblastoma variant, not merely a globally more unstable GBM subset. The finding is small-cohort evidence, but it is strong enough to argue that gliosarcoma should be stratified rather than silently pooled in molecular studies and clinical trials.

Vineet Reddy

Relapsed pediatric B-cell acute lymphoblastic leukemia remains clinically hard because recurrence is not only a return of leukemia burden, but a selected biological state. We reanalyzed 49 paired diagnosis-relapse marrow expression profiles from GEO series GSE28460 and challenged the resulting relapse program in the independent diagnostic cohort GSE7440. Across 21,755 annotated genes in GSE28460, 505 genes changed at relapse at FDR < 0.10 with an absolute mean paired log2 fold-change above 0.35. The relapse program was dominated by cell-cycle and DNA-replication biology, with cell division, sister chromatid cohesion, mitotic nuclear division, DNA replication, and G1/S transition as the strongest enriched processes. Curated signature analysis recovered increased cell-cycle scoring and decreased glucocorticoid-response scoring at relapse. In 59 GSE7440 diagnosis samples with outcome labels, the top relapse-up minus relapse-down program did not distinguish later relapse from complete continuous remission (AUC 0.48; Mann-Whitney p = 0.80). This negative external test sharpens the interpretation: the signal is a relapse-state program, not a baseline bulk prognostic signature.

Vineet Reddy

Peritumoral FLAIR hyperintensity is clinically important in meningioma, but it is often treated as a qualitative companion to tumor size rather than as a quantitatively separable imaging phenotype. We analyzed 1,000 public BraTS 2023 pre-operative meningioma training segmentations to measure enhancing tumor, non-enhancing or necrotic core, and peritumoral FLAIR hyperintensity volumes from expert-refined masks. Edema was present in 536 cases. Median segmented tumor volume was 9.21 ml (IQR 2.08-31.17), while median edema volume was 0.08 ml (IQR 0.00-23.54), reflecting a sharply zero-inflated and right-tailed distribution. Among cases with nonzero edema, log edema volume increased with log tumor volume (Spearman rho = 0.47, p = 3.6e-30), and a heteroskedasticity-robust log-log model estimated an approximately proportional scaling exponent of 0.97 (95% CI 0.80-1.14). However, tumor volume explained only 21.6% of edema-volume variance. The 90th percentile edema-to-tumor ratio was 3.87; these high-ratio cases had far greater edema burden despite smaller median tumor volume than the remaining edema-positive cohort. These findings argue that meningioma-associated FLAIR hyperintensity should not be reduced to tumor size. Simple compartment volumes provide a useful baseline, but edema burden appears to encode additional biology, anatomy, or acquisition-context effects that require clinical metadata and richer imaging models.

Vineet Reddy

Neuroligin 3 (NLGN3) sits at an awkward intersection: it is an autism-linked synaptic adhesion molecule, a shed glioma mitogen, and an astroglial perturbation whose transcriptional consequences are still being defined. We asked whether these disease contexts share a measurable NLGN3-responsive transcriptional axis. We reanalyzed two public human RNA-seq datasets: recombinant NLGN3 exposure in SU-pcGBM2 high-grade glioma cultures (GSE99045) and NLGN3 R451C knock-in astroglia-enriched organoids versus isogenic 3D wild type (GSE283484). Because the glioma dataset provides FPKM rather than raw counts, we used effect-size and rank-based convergence rather than count-model differential expression. Global transcriptome concordance was weak (Spearman rho = 0.014, p = 0.109 across 12,598 matched expressed genes), but genes with moderate effects in both systems were same-direction more often than expected (905/1,677, binomial p = 6.3e-4). Top upregulated genes showed strong same-direction overlap: the top 1,000 induced genes overlapped by 198 genes versus 79 expected (odds ratio = 3.32, Fisher p = 2.45e-36). Curated module analysis identified extracellular matrix and adhesion genes as the clearest shared signal (mean log2 fold-change +0.15 in glioma and +1.80 in astroglia; empirical product FDR = 0.0064), not canonical neurexin-neuroligin synaptic genes. These results argue that cross-context NLGN3 biology is not a generic synaptic transcriptome program. It is a focused adhesion-remodeling response that appears in both glioma exposure and mutant human astroglia.

Vineet Reddy

Neuroligin-1 (NLGN1) is often discussed through two narrow lenses: extracellular neurexin-binding biology and autism-associated clinical variation. That framing suggests an attractive hypothesis: clinically submitted NLGN1 protein variants might concentrate in splice-regulated or extracellular adhesion regions that tune neuroligin-neurexin function. We tested that hypothesis directly using public variant and protein-annotation resources. ClinVar contained 169 NLGN1 records, including 43 protein-change submissions that could be mapped unambiguously to the UniProt Q8N2Q7 coordinate system by reference-amino-acid matching. As a background catalogue, we mapped 603 Ensembl/dbSNP missense variants to the same protein coordinate system using Ensembl VEP. We then compared clinical and background variant localization across the extracellular domain, cholinesterase-like core, splice-site A segment, glycosylation/disulfide-proximal windows, juxtamembrane segment, transmembrane helix, and cytoplasmic tail. The public data did not support a simple clinical hotspot model. ClinVar mapped variants were not enriched in the extracellular domain, cholinesterase-like core, splice-site A segment, or post-translational/structural windows relative to dbSNP missense variants. Although ClinVar variants were closer to splice-site A by median distance, only one mapped variant lay inside the A segment. UniProt-curated functional natural variants also spanned both extracellular and cytoplasmic positions. These results argue that current public clinical submissions are too sparse and VUS-heavy to justify a splice-site-centered clinical interpretation of NLGN1. Mechanistic claims about NLGN1 variants should remain variant-specific until patient-level genotype-phenotype data or functional assays provide stronger evidence.

Vineet Reddy

Urban floras are often described as lists, but lists do not show how urban structure partitions native and introduced plant signals. I analyzed 12,137 research-grade iNaturalist plant observations from Berkeley, California, spanning 669 observed taxa, and compared taxa flagged as introduced by iNaturalist across four fixed west-to-east spatial strata: west flatlands, central urban grid, campus/creek corridor, and hill/canyon edge. Equal-effort resampling used the smallest stratum size, 102 observations, with 1,000 bootstrap draws per stratum. The strongest pattern was not a smooth urban-to-wildland gradient. The central urban grid had the highest introduced observation share (0.782; 95% bootstrap interval 0.696-0.853), while the hill/canyon edge had the lowest share (0.451) but was sparsely sampled. The campus/creek corridor was intermediate (0.639; 0.549-0.726), close to the west flatlands (0.616; 0.520-0.706), despite Berkeley's documented Strawberry Creek restoration history. These results suggest that Berkeley's plant records preserve a clear introduced-species core in the urban grid, but the available citizen-science data do not yet support a strong claim that the campus creek corridor has separated from the broader urban introduced-plant signal.

Vineet Reddy

Rare gastrointestinal disorders are often described as "emerging," but the term is usually used loosely. We tested whether non-esophageal eosinophilic gastrointestinal disorders (EGIDs) show a measurable emergence signal in public biomedical data and whether public gastric-biopsy transcriptomics support treating eosinophilic gastritis as a distinct disease focus rather than a derivative of eosinophilic esophagitis. We queried PubMed yearly record counts from 2010 through 2025 for eosinophilic gastritis, gastroenteritis, colitis, enteritis/duodenitis, eosinophilic esophagitis, and comparator rare gastrointestinal disorders, and we filtered ClinicalTrials.gov records by exact condition names. We then reanalyzed GEO GSE54043, a gastric antrum microarray dataset containing five eosinophilic gastritis cases and five controls. Eosinophilic gastritis had the largest publication growth signal: mean annual PubMed records increased from 4.4 in 2010-2014 to 28.4 in 2021-2025, a 5.90-fold increase. Eosinophilic enteritis/duodenitis and eosinophilic colitis also exceeded the eosinophilic esophagitis comparator on relative growth, while their exact-condition trial footprints remained small. In GSE54043, simple Welch-test reanalysis was underpowered after false-discovery correction, but the largest nominal signals included eosinophil-associated and gastric inflammatory transcripts, consistent with prior mechanistic work. The strongest defensible conclusion is not that a new disease has appeared. It is that non-esophageal EGIDs, especially eosinophilic gastritis, are moving from scattered rarity toward an identifiable research category faster than several other rare gastrointestinal comparators, while still lacking the clinical-trial density and public molecular replication that would make the field mature.

Vineet Reddy

NLGN1 is a synaptic adhesion gene with established roles at excitatory synapses and repeated links to neurodevelopmental disease, but public discussion often collapses it into a vague claim that it is simply “brain expressed.” We tested whether that simplification survives direct inspection of public human transcriptomic atlases and disease-relevant cortical cohorts. Using the BrainSpan developmental transcriptome together with adult regional and single-nucleus brain RNA profiles from the Human Protein Atlas, we quantified NLGN1 across developmental stages, adult subregions, and cell-type groupings. We then examined NLGN1 in independent schizophrenia and autism dorsolateral prefrontal cortex RNA-seq cohorts and in a 63-donor schizophrenia Visium DLPFC spatial transcriptomics cohort. The resulting picture was not a simple forebrain excitatory trajectory. In BrainSpan, median NLGN1 expression was higher in prenatal than postnatal samples across the cortical, hippocampal, and amygdalar groupings we could compare. In contrast, adult regional expression was strongest in cerebellar cortex, vermis, and flocculonodular lobe. Single-nucleus summaries showed strong neuronal expression but also an unexpectedly prominent oligodendrocyte-precursor signal, while vascular and immune compartments remained low. Despite NLGN1’s disease relevance, neither bulk DLPFC cohort nor the layer-resolved Visium cohort showed a strong case-control shift.

Vineet Reddy

Venture investment in healthtech has not moved as a simple boom-bust story. Using a hand-compiled public aggregate series of U.S. digital health venture funding from 2011 through 2025, triangulated against CB Insights, SVB, and NVCA/PitchBook market reports, this paper identifies five funding regimes: category formation, pre-pandemic scale-up, pandemic boom, post-2021 correction, and selective rebound. Funding rose from roughly $0.9B in 2011 to $29.1B in 2021, then fell 63.2% by 2023. By 2025, funding had recovered to $14.2B, but deal count remained far below the 2021 peak and even declined relative to 2023. The post-correction market is therefore not a broad reopening. It is a concentration regime: fewer companies receive larger checks, mega-rounds matter again, and investor attention has shifted toward AI-enabled infrastructure, provider operations, and businesses with clearer evidence of revenue quality or exit relevance. The cycle that emerged is not merely periodic enthusiasm for healthcare technology. It is a repeated migration between access expansion and discipline, with each correction preserving some infrastructure from the prior boom while raising the burden of proof for the next cohort.

Vineet Reddy

Public pediatric brain tumor datasets now make pan-histology molecular prognostic analyses possible, but that breadth creates a hard statistical problem: tumor class, molecular burden, treatment, and outcome are not exchangeable across diagnoses. We reanalyzed public summary files from OpenPBTA release v23, merging harmonized clinical annotations with coding tumor mutation burden for 782 survival-annotated molecular profiles from 943 enrolled participants. Hypermutated samples, defined as coding tumor mutation burden at or above 10 mutations/Mb, had worse unadjusted overall survival than lower-burden samples (log-rank p=0.0036), but the hypermutated group was small (n=7) and diagnosis-mixed. Mutation burden also varied strongly across broad histology classes. After residualizing log-transformed mutation burden against age at diagnosis and broad histology, the association with observed death weakened (Mann-Whitney p=0.089). Within the three broad histology strata large enough for sensitivity checks, median-split TMB did not show a consistent high-burden survival penalty. These results do not support mutation burden as a diagnosis-independent prognostic marker in this public summary-data setting. They instead quantify a useful failure mode: pan-pediatric CNS tumor models can rediscover diagnosis composition unless diagnosis structure is modeled explicitly. OpenPBTA remains an unusually valuable resource, but survival modeling from public summaries should be framed as hypothesis generation unless treatment, molecular subtype, sampling phase, and diagnosis-specific effects are incorporated.

Vineet Reddy

Open neuroimaging datasets can accelerate trigeminal neuralgia research only if their public clinical metadata are complete enough to reproduce outcome analyses. We audited the public OpenNeuro ds005713 release and tested whether its accessible tabular data support a common prognostic question: does coded Sindou neurovascular compression grade predict postoperative pain? The release README describes 119 patients, 53 healthy controls, 64 follow-up MRI records, and 93 telephone follow-ups. In contrast, the participants.tsv file fetched from the public S3 endpoint on April 25, 2026 contained 51 rows, all trigeminal-neuralgia patients, while MRIQC group files contained 110-158 unique participant IDs depending on modality. This mismatch makes imaging availability much broader than public clinical-outcome availability. Within the 51-row clinical table, 49 patients were operated, but immediate postoperative outcome was strongly ceilinged: 42 of 49 operated patients (85.7%) had 100% pain reduction. Second follow-up pain scores were available for 46 operated patients, and only 5 had residual pain greater than zero. Sindou grade showed no stable association with follow-up pain (Spearman rho -0.10, 95% bootstrap interval -0.46 to 0.28), Fisher's exact p = 0.50 after dichotomization, and leave-one-out multivariable prediction performed poorly (AUC 0.44 with missing Sindou retained; AUC 0.32 among known Sindou cases). The result is not that neurovascular compression is unimportant. It is sharper: this public release, as currently auditable from its tabular metadata, is not yet sufficient for reproducible surgical prognostic modeling.

Vineet Reddy

Large Balmer decrements in optical galaxy spectra are usually treated as dust corrections. This paper asks whether the extreme tail is more interesting: a compact way to find dust-buried narrow-line active galactic nuclei in SDSS spectroscopy. I queried the SDSS DR17 SkyServer catalog for 300,893 low-redshift galaxy spectra with robust Halpha, Hbeta, [O III] 5007, and [N II] 6584 detections, measured Halpha/Hbeta, and classified each source on the BPT diagram. Galaxies with Halpha/Hbeta >= 8 are rare in this quality-selected sample (1,054 objects), but 84.7% of them fall in the composite or AGN region, compared with 24.0% of redshift-matched controls with 3 <= Halpha/Hbeta < 6. A Fisher exact test gives an odds ratio of 17.6 and p = 8.7e-185. The active fraction also rises monotonically across Balmer-decrement bins, from 18.3% at 2.8-4 to 84.7% above 8. The result does not prove that every high-decrement object hosts an obscured nucleus; aperture effects, stellar absorption residuals, and metallicity structure still matter. But the effect is too large to dismiss as a smooth dust tail. Extreme Balmer decrements are a useful, cheap triage flag for hidden nuclear activity in SDSS-like optical spectra.

Vineet Reddy

Wafer-scale AI processors are usually described as GPU replacements because they integrate an entire silicon wafer into one accelerator. That framing hides the more useful question: which transformer inference workloads actually fit the wafer-scale design point? We combine public accelerator specifications for Cerebras WSE-3, NVIDIA A100, NVIDIA H100, NVIDIA DGX B200, and AMD MI300X with representative transformer KV-cache sizes, a spill-sensitivity model, and an inspection of the open WaferLLM benchmark configuration grid. WSE-3 occupies an extreme bandwidth-density point: 21 TB/s of local SRAM bandwidth over 44 GB, or 0.477 TB/s per GB, roughly an order of magnitude above the GPU points in this comparison. The same design gives up local capacity. Under BF16 KV-cache assumptions, WSE-3 can hold about 90k active 7B-class tokens, 18k 70B-class tokens, and 5.7k 405B-class tokens before spilling beyond on-wafer storage. With only half of local memory available for 70B-class KV cache and spilled bytes moving at 10 percent of local bandwidth, the modeled WSE-3/H100 memory-time advantage persists for small active-token products but crosses a parity boundary as batch and context grow. WaferLLM's public WSE-3 decode configs are consistent with this interpretation: their single-layer kernel KV allocations fit comfortably on wafer, while full-model KV residency would be a different constraint. The result is not a generic accelerator ranking. It is a phase boundary, and it argues that wafer-scale systems should be evaluated by locality envelopes, not only peak FLOP counts.

Vineet Reddy

Oropouche virus disease is being treated as an emerging arboviral threat in the Americas, but the public-health question is not simply whether reported case counts are rising. The harder question is whether recent detection follows the established dengue surveillance footprint or reveals a different epidemic geography. We compiled PAHO/WHO Oropouche case counts for 2024 and 2025 through epidemiological week 30, PAHO dengue situation-report counts through epidemiological week 25 of 2025, and Brazil state-level Oropouche counts from the August 2025 PAHO update. The analysis is descriptive, because public weekly Oropouche line lists were not available. The signal is still sharp. Autochthonous Oropouche reports in 2025 remained Brazil-dominated: Brazil accounted for 93.0% of reported regional autochthonous cases. Yet within Brazil, 98.8% of the state-level burden in the PAHO update occurred in states classified here as outside the historic Amazon-centered frame, with Espirito Santo and Rio de Janeiro alone accounting for 74.2% of reported Brazilian cases. Panama also moved from 16 reported autochthonous cases in 2024 to 501 in 2025. A dengue comparator showed that Oropouche detection is not a simple function of dengue burden: Panama's Oropouche-to-dengue ratio was far higher than Brazil's, while several high-dengue countries had no reported Oropouche signal in the compiled update. These findings argue for targeted Oropouche testing in arboviral syndromic surveillance, especially where dengue-like illness is common but dengue-confirmed fractions, travel links, or unusual neurologic or pregnancy outcomes are discordant.

Vineet Reddy

State restrictions on municipal broadband sit at the center of a live U.S. infrastructure controversy: supporters frame them as protection against risky public ventures, while critics frame them as incumbent protection that limits local responses to broadband market failure. Prior county-panel research found that municipal and cooperative restrictions were associated with lower broadband availability before the FCC's Broadband Data Collection replaced Form 477 availability data. This paper asks whether the controversy is visible in the current post-Form-477 county data. I merged the FCC/Esri June 2024 county Broadband Data Collection layer with 2023 ACS county controls, Census land-area data, and a 2024 Institute for Local Self-Reliance coding of 16 states with municipal-broadband preemption laws. Across 3,133 counties and county equivalents, preemption-state counties had lower BSL-weighted served shares than other counties (92.8% versus 93.9%) and higher BEAD-eligible shares (7.2% versus 6.1%). In BSL-weighted county regressions controlling for population density, income, poverty, education, county BSL scale, and Census division, preemption states remained associated with a 1.6 percentage-point lower served share and a corresponding 1.6 percentage-point higher BEAD-eligible share. The adjusted associations for fiber service, unique provider count, and household broadband subscription were negative but imprecise. The evidence therefore supports a narrow claim: current BDC county data show a modest availability disadvantage in preemption states, but they do not settle the broader controversy over municipal broadband competition.

Vineet Reddy