Prader-Willi syndrome arises from errors in the expression of paternally inherited genes on chromosome 15, affecting roughly one in ten thousand to one in thirty thousand births. Infants with the condition typically show weak muscle tone, a poor suck that makes feeding difficult, and mild facial differences such as a narrow forehead, almond-shaped eyes, and a downturned mouth. As childhood progresses, the picture changes dramatically: an unrelenting drive to eat emerges, leading to life-threatening obesity if not managed, alongside cognitive and behavioral difficulties. Early recognition matters enormously, because growth hormone therapy, feeding support, and structured developmental care initiated soon after birth can measurably improve outcomes. The catch is that the neonatal facial phenotype, the very signal clinicians might rely on, is faint and easily overlooked in a squirming, sleeping infant.

The Fudan team, led by Yunqian Zhu and Wenhao Zhou, set out to build a diagnostic aid that works precisely where existing tools falter. Their strategy rested on a pragmatic observation: photographs of older children with the syndrome are relatively abundant, while images of newborns are scarce and expensive to collect. Could a model trained on the faces of older children still detect the syndrome in newborns whose facial features are far less pronounced? To find out, the researchers assembled a training cohort of 113 children drawn from a public platform, of whom 21 carried confirmed diagnoses of Prader-Willi syndrome, and a separate validation cohort of newborns at their own hospital consisting of ten affected infants and thirteen controls.

The technical approach combined multiple feature sets and learning algorithms. The team extracted a detailed panel of facial measurements, including distances, angles, and shape descriptors capturing everything from eye spacing to lip contour, alongside age. They then trained six different classification algorithms, spanning from logistic regression to Random Forests and AdaBoost ensembles, on three versions of the feature set: the complete panel, a top-ten subset of the most informative features, and a clinically selected subset chosen for relevance to the syndrome. Performance was judged using the area under the precision-recall curve, a metric well suited to datasets where positive cases, like rare diseases, are outnumbered by controls. Across all feature sets, the Random Forest and AdaBoost models emerged as the strongest performers.

The headline result came from the Random Forest trained on the complete feature set. In cross-validated testing on the training cohort, it achieved a mean sensitivity of 0.97, with a 95 percent confidence interval stretching from 0.84 to 1.00, and a specificity of 0.95, correctly identifying nearly every child with the syndrome while rejecting nearly all children without it. The model was also well calibrated, meaning its predicted probabilities tracked actual likelihoods rather than being systematically overconfident or timid, a property the researchers emphasize as essential for any tool intended to support real clinical decisions rather than inflate benchmark scores.

Peering inside the model revealed a face in numbers. Age, face width, and features of the upper lip carried the greatest predictive weight, consistent with the well-described dysmorphic profile of the syndrome. More intriguingly, the analysis surfaced four facial features not previously reported in the context of Prader-Willi syndrome: jaw symmetry, chin shape, nose width, and jaw width. A sensitivity analysis showed that 26 facial features contributed more to classification than age alone, suggesting that the model was genuinely reading facial structure rather than simply learning that children with the condition tend to be photographed at certain ages. These newly identified dimensions of the syndrome’s facial signature could inform clinicians and researchers alike, expanding what is understood about how a single genetic lesion reshapes craniofacial development.

The decisive test, however, was neonatal. When the researchers applied their child-trained model to the newborn validation cohort, they first had to adjust the decision threshold, because the probability scores that fit older children needed recalibration for the subtler neonatal face. After this re-estimation, the Random Forest model detected 80 percent of affected newborns while correctly clearing 77 percent of controls. To benchmark this performance, the team compared it against Face2Gene, a widely used commercial facial analysis app for genetic syndromes. On the same newborn images, Face2Gene achieved only 70 percent sensitivity and a mere 15 percent specificity, meaning it flagged almost every baby as suspect. At a false positive rate of 0.25 or less, the Fudan model delivered comparable sensitivity but dramatically higher specificity, a distinction that matters in a screening context where alarmingly frequent false alarms would erode trust and swamp follow-up genetic testing capacity.

Beyond the immediate clinical promise, the study carries a broader methodological message for the field of computational phenotyping. Artificial intelligence tools for rare disease detection, including deep learning systems such as DeepGestalt and GestaltMatcher, have demonstrated impressive diagnostic power in clinics and genetic consultation settings, but their training data overwhelmingly feature older children and adults. Newborns, with their rounded faces, abundant subcutaneous fat, and physiologically underdeveloped craniofacial structures, represent a distribution shift that most models handle poorly. The Fudan results suggest a practical workaround: rather than waiting years to accrue a newborn dataset of sufficient size, researchers can train on pediatric data and transfer the learned phenotype knowledge to neonates through careful threshold adjustment. That could accelerate screening tools for a whole family of genetic conditions where the newborn window of intervention is narrow and the data pipeline begins nearly empty.

Caveats remain, and the authors are candid about them. The validation cohort numbered just 23 newborns, a pilot sample sufficient to demonstrate feasibility but far short of what a population-level screening program would demand. The control group’s size and composition limit how confidently the specificity estimates can be generalized, and the model must eventually prove itself across diverse ethnic populations, camera conditions, and gestational ages. The researchers also note that facial analysis is intended as a triage tool, not a replacement for the methylation testing that definitively diagnoses the syndrome. A positive screen would simply channel scarce genetic testing resources toward the babies most likely to benefit, shortening what can otherwise be a diagnostic odyssey measured in months or years.

Still, the trajectory of the work is hard to ignore. With a new drug recently approved to treat hyperphagia in older patients and advances such as hypothalamic organoids offering laboratory models of the underlying biology, the therapeutic landscape for Prader-Willi syndrome is shifting, and the value of catching the condition at birth rises in step. A camera and an algorithm cannot replace a pediatric geneticist, but they can stand watch over every delivery ward, quietly and cheaply, flagging the one baby in ten thousand whose face holds a message that human eyes, on their own, too often miss. If larger prospective studies replicate the Shanghai results, facial phenotype screening could become the first step of a standard newborn pathway, ensuring that children with Prader-Willi syndrome receive growth hormone, feeding support, and early intervention before the syndrome’s most damaging consequences take root.

Subject of Research: Machine learning-based facial phenotype screening for early detection of Prader-Willi syndrome in newborns

Article Title: Early identification of Prader-Willi syndrome in newborns using facial phenotype-based machine learning

Article References: Zhu, Y., Chen, H., Chen, J., Yang, L., Dong, X., Ni, Q., & Zhou, W. (2026). Early identification of Prader-Willi syndrome in newborns using facial phenotype-based machine learning. Pediatric Research. https://doi.org/10.1038/s41390-026-05491-6

Image Credits: AI Generated

DOI: 10.1038/s41390-026-05491-6

Keywords: Prader-Willi syndrome, machine learning, facial phenotype, newborn screening, rare disease, random forest, Face2Gene, genetic diagnosis, dysmorphism, pediatric research, artificial intelligence, early diagnosis

Central precocious puberty, or CPP, occurs when the hypothalamic-pituitary-gonadal axis is activated prematurely, leading to early breast development in girls and virilization in boys before the customary age. The condition matters clinically for more than cosmetic reasons: early activation of the axis accelerates skeletal maturation, can compromise final adult height, and may carry psychological and social consequences for children whose bodies mature years ahead of their peers. Over the past decade, loss-of-function mutations in MKRN3, an imprinted gene located in the Prader–Willi syndrome region of chromosome 15, have emerged as one of the most common genetic causes of familial CPP. Yet the same genomic neighborhood presents a paradox: when MKRN3 is deleted as part of the larger deletion that causes Prader–Willi syndrome, patients typically develop hypogonadism and delayed or incomplete puberty rather than precocious puberty. Because MKRN3 is paternally expressed and maternally imprinted, its loss through mutation removes an inhibitory brake on gonadotropin-releasing hormone secretion, hastening puberty, whereas the complex multi-gene deletion in Prader–Willi syndrome disrupts hypothalamic development in ways that blunt reproductive function.

The mechanistic basis for these opposing reproductive phenotypes has remained largely unclear. To address this, the team led by Yu-Yu Jin, Xiao Wang, Lin Yang, Jian Mu, and Fei-Hong Luo performed whole-exome sequencing on 98 Chinese children with central precocious puberty. They then carried out a systematic review of previously reported MKRN3 pathogenic and likely pathogenic variants to summarize genotype–phenotype correlations. Finally, they performed genome-wide DNA methylation profiling in CPP patients carrying the MKRN3 pathogenic variant and compared their methylation patterns with those of patients with Prader–Willi syndrome, patients with idiopathic central precocious puberty, and healthy controls. This three-pronged design allowed the investigators to move from variant discovery to clinical correlation to epigenetic characterization within a single cohort framework.

The sequencing effort identified a pathogenic frameshift variant, designated c.476dupC (p.Ala159fs*15), which introduces a duplication of a single cytosine base that shifts the reading frame and generates a premature stop codon. Notably, this is the first frameshift mutation reported within the inter-C3H1 hotspot region of MKRN3 in an Asian cohort, a finding that further confirms the functional significance of this segment of the protein. The MKRN3 protein carries multiple C3H-type zinc finger domains and a RING finger domain, and pathogenic variants have previously clustered in certain regions, making variant location an important clue to functional impact. The RING finger domain confers E3 ubiquitin ligase activity, meaning the protein tags other cellular proteins for degradation or modification, while the zinc finger domains mediate binding to RNA and protein targets. Frameshift variants that truncate the protein before these domains are complete are therefore expected to eliminate function almost entirely.

The genotype–phenotype analysis revealed a striking correlation between variant severity and clinical presentation. Patients carrying severe MKRN3 variants, such as frameshift or nonsense mutations predicted to truncate the protein, exhibited significantly earlier pubertal onset than those with missense mutations, beginning at a mean age of 5.80 years compared with 7.50 years (P = 0.029). The severity of the hormonal activation also differed: children with severe variants showed significantly higher luteinizing hormone peak levels during gonadotropin-releasing hormone stimulation testing, at 34.55 versus 11.00 IU/L (P = 0.047). These findings suggest that the degree of residual MKRN3 function, or the manner in which a truncated protein is produced and degraded, influences how vigorously the reproductive axis is disinhibited. In practical terms, the two-and-a-half-year difference in mean onset age between the variant classes is clinically meaningful, since earlier activation leaves a longer window of compromised growth potential and often requires more aggressive intervention with gonadotropin-releasing hormone agonist therapy to pause pubertal progression.

The most novel component of the study was the methylation analysis. DNA methylation, the addition of methyl groups to cytosine bases across the genome, is a key epigenetic mechanism that regulates gene expression without altering the underlying DNA sequence. Epigenetic regulation is known to be central to the timing of puberty: the reactivation of gonadotropin-releasing hormone secretion at the end of childhood is thought to depend on carefully orchestrated changes in chromatin state within hypothalamic neurons, and animal studies have shown that disrupting epigenetic machinery can shift pubertal timing substantially. MKRN3 itself has been implicated in epigenetic switching through its ubiquitination of MBD3, a component of chromatin-remodeling complexes. Given that MKRN3 sits in an imprinted genomic domain governed by methylation marks, the investigators asked whether MKRN3-associated CPP carries a distinct methylation signature.

Surprisingly, the analysis revealed no differences in methylation at the MKRN3 locus itself between the study groups. However, the comparison between MKRN3-CPP patients and Prader–Willi syndrome patients identified 18,609 differentially methylated positions across the genome, indicating that the two conditions, despite sharing the same chromosomal region, are epigenetically worlds apart. Key findings included pronounced hypermethylation of IGSF10 (a difference in methylation beta values of 0.37), ZC3H18 (0.27), SH3RF3 (0.36), and PTH1R (0.28) in the MKRN3-CPP group relative to PWS, alongside hypomethylation of MAGEL2 (−0.19) and PTPA (−0.23), where the delta-beta value represents the difference in DNA methylation beta values between the groups.

Several of these differentially methylated genes have plausible links to reproductive biology. IGSF10 is well established as a gene involved in the migration of gonadotropin-releasing hormone neurons during embryonic development, and mutations in it cause delayed puberty, making its epigenetic dysregulation particularly relevant to pubertal timing disorders. PTH1R, the parathyroid hormone 1 receptor, signals through the cAMP/protein kinase A pathway and is best known for its actions on bone. MAGEL2, which lies within the Prader–Willi syndrome critical region, is required for the production of secretory granules and neuropeptides in the hypothalamus, and its loss contributes to the impaired melanocortin pathway function characteristic of PWS. PTPA, a protein phosphatase 2A activator, has been shown to physically associate with the C-terminus of GPR54, the kisspeptin receptor that is a master gatekeeper of puberty, and has roles in osteoblast differentiation. SH3RF3 has been connected to the LIN28B pathway, which itself is among the genetic loci associated with age at menarche in large genome-wide association studies. Taken together, these targets trace a coherent map of the neuroendocrine and developmental circuits that determine when and how forcefully the reproductive axis awakens.

The authors propose that this pattern of differential methylation in downstream neuroendocrine pathways could potentially explain how divergent pubertal phenotypes arise in the setting of MKRN3 deficiency. In isolated MKRN3 mutations, the loss of the gene’s inhibitory action in kisspeptin-expressing neurons, together with its known role in modulating the stability and translation of GNRH1 messenger RNA through ubiquitination of poly(A)-binding proteins, removes the pubertal brake. In Prader–Willi syndrome, the methylation landscape of genes such as MAGEL2, a paternally expressed neighbor of MKRN3, is altered in the opposite direction, and the broader deletion disrupts neuropeptide production and hypothalamic circuitry, leading to insufficient rather than excessive gonadotropin secretion. The methylation data thus suggest that the clinical contrast between the two conditions may be inscribed in the epigenetic programming of shared downstream pathways rather than in MKRN3 itself. If confirmed, this would represent a conceptual shift: the direction of the pubertal phenotype would be determined not by the presence or absence of a single gene, but by the constellation of epigenetic states across the network it regulates.

The study has limitations worth noting. The methylation profiling was performed in peripheral blood samples, which may not fully reflect methylation states in the hypothalamic neurons that actually govern puberty, a common constraint in human neuroendocrine epigenetics. Blood-based methylation signatures can be confounded by differences in cell-type composition between patient groups, and tissue specificity remains a persistent challenge for the field. The cohort of MKRN3-mutant patients available for methylation comparison was necessarily small given the rarity of the condition, and the cross-sectional design cannot establish whether the methylation differences are causes or consequences of the divergent pubertal phenotypes. The researchers also note that the data were processed using beta-mixture quantile normalization to correct probe design bias in Illumina methylation arrays, and variant interpretation followed the American College of Medical Genetics and Genomics standards, but functional validation of the identified methylation changes will require further work, ideally in neuronal models or longitudinal cohorts.

Nevertheless, the implications are considerable. For clinicians, the genotype–phenotype correlation reinforces the value of molecular testing in children with central precocious puberty, particularly those with a family history or very early onset, and suggests that children with truncating MKRN3 variants may warrant particularly close attention to the pace of pubertal progression. Because MKRN3-related CPP is inherited in a pattern determined by the imprinting of the gene, molecular diagnosis also carries implications for genetic counseling of affected families, including the identification of relatives who carry the variant but may be asymptomatic. For researchers, the identification of an epigenetic signature distinguishing MKRN3-CPP from Prader–Willi syndrome opens a line of inquiry into how ubiquitin ligases and chromatin modifiers in the imprinted 15q11-q13 domain jointly sculpt the timing of human puberty. The work was supported by the National Natural Science Foundation of China and the Science and Technology Commission of Shanghai Municipality, and the data are available from the corresponding author on request.

More broadly, the study adds to a growing body of evidence that epigenetic marks can serve as interpretable intermediates between genetic lesions and complex developmental phenotypes. Just as methylation scores are being explored as predictors of metabolic disease risk, methylation profiling of pubertal disorders may eventually help stratify patients, refine prognosis, and illuminate therapeutic targets in the neuroendocrine circuitry that launches human reproduction. In an era when sequencing routinely uncovers rare variants of uncertain significance, studies that connect variant class, clinical severity, and epigenetic state provide exactly the kind of layered evidence needed to translate genomic findings into meaningful care for children whose puberty arrives, or fails to arrive, at the wrong time.