Yousheng Yan

Background: Joubert syndrome (JS) is a genetically heterogeneous ciliopathy wherein non-canonical splice-site (NCSS) variants are frequently reported as variants of uncertain significance (VUS), complicating prenatal counseling. We aimed to clarify the clinical significance of an NCSS in CC2D2A through functional validation. Methods: Trio-WES was performed on a 13-year-old female proband with clinically diagnosed JS. Candidate variants were validated by Sanger sequencing and familial segregation. An in vitro minigene assay and RT-PCR on patient peripheral blood RNA were used to assess the splicing effect of c.439-5T>G. Variant classification followed the ACMG/AMP guidelines using the ClinGen quantitative Bayesian framework. Results: Compound heterozygous CC2D2A variants were identified: c.439-5T>G (maternal) and c.4732G>A (p.(Val1578Met), paternal); both were initially classified as VUS. Minigene analysis showed that c.439-5T>G caused aberrant splicing, retaining the intronic nucleotides c.439-4_439-1 (ATAG) at the exon 6–7 junction (r.438_439insauag) and introducing a frameshift and premature termination (p.(Pro147IlefsTer14)). Patient RNA confirmed this pattern. c.439-5T>G was reclassified as likely pathogenic (PVS1_VeryStrong + PM2_Supporting). c.4732G>A remained a VUS (PM3 + PM2_Supporting + PP3) on an independent allele basis; its trans configuration with the likely pathogenic allele supported a family-specific biallelic explanation but did not establish the pathogenicity of the missense allele. The fetus in a subsequent pregnancy was confirmed to carry both variants, establishing the compound heterozygous state. Conclusions: Minigene-based validation provided evidence to reclassify a CC2D2A NCSS from VUS to likely pathogenic, enabling variant-directed prenatal testing and informed reproductive counseling in this JS family. The paternal missense variant c.4732G>A (p.(Val1578Met)) remains a VUS; its trans configuration with the likely pathogenic allele supports, but does not establish, a biallelic explanation. This study illustrates how functional assays can clarify splicing variants that in silico predictions alone cannot reliably interpret.

Background: Phenotype-driven prenatal diagnosis relies on the precise correlation between ultrasound findings and genetic outcomes; however, this process is hindered by the unstructured nature of clinical ultrasound reports. While large language models (LLMs) hold the potential to address this challenge, their specific application in this domain remains systematically underexplored. Objective: To establish an effective LLM implementation framework for the clinical multidimensional classification of prenatal ultrasound reports, we evaluated the open-source DeepSeek-V3.2 family on real-world anomalous reports—covering both factual and subjective categories—while integrating retrieval-augmented generation (RAG) and chain-of-thought (CoT) reasoning. Methods: From a cohort of 4256 pregnancies, we extracted 254 reports with fetal anomalies. We comprehensively evaluated both the high-speed base model (DeepSeek-V3.2-B) and the reasoning-enhanced model (DeepSeek-V3.2-R) across all 5 classification dimensions, comprising 4 factual extraction tasks—primary classification, standardized terminology, anatomical system, and abnormality count—and 1 subjective severity assessment. We further explicitly evaluated the efficacy of RAG for the subjective tasks. Finally, to validate the clinical utility of this approach, we performed a correlation analysis between the expert-validated multidimensional phenotypic profiles and definitive genetic outcomes derived from amniocentesis. Results: While V3.2-B achieved high efficiency in factual tasks (accuracy and F ₁ -score >90%), it underperformed in subjective severity grading (56.6% accuracy), exhibiting a recall of 0 for minor anomalies. Crucially, while RAG significantly improved both models’ performance on internal retrieval datasets ( P <.05), this benefit did not generalize to external test datasets ( P >.05). In contrast, the V3.2-R model utilizing CoT reasoning achieved superior robustness (86% accuracy and F ₁ -score=0.75) on external data without RAG; notably, introducing RAG to V3.2-R degraded performance to 81%, suggesting potential noise interference. Clinical validation against amniocentesis outcomes confirmed that accurate multidimensional phenotypic profiles significantly stratified pathogenic genetic risks. Conclusions: The rapid base models are efficient for factual classification, and RAG enhances performance on data similar to the knowledge base, whereas CoT is indispensable for subjective assessment. Within the constraints of our dataset and current retrieval implementation, CoT proved more robust than RAG for subjective assessment. However, this finding is specifically tied to our experimental setup and should not be generalized as a universal conclusion. We recommend clinically adopting this adaptive “fast-slow” LLM framework to efficiently perform the multidimensional classification of prenatal ultrasound anomalies. This privacy-preserving, locally deployable solution provides a scalable path to accelerate phenotype-genotype research and optimize invasive diagnostic decision-making.

The open-source reasoning large language model DeepSeek-R1 is increasingly being used in hospitals, but its multiple parameter versions, especially the distilled models, have not been fully evaluated for diagnostic performance. To address this, paired comparisons were conducted using five DeepSeek-R1 models and their respective base models. The models were tested on a diagnostic dataset of 110 simulated clinical cases from open access data, covering internal medicine, surgery, neurology, gynecology, and pediatrics, and categorized by incidence (frequent, less frequent, rare). The models were tasked with generating five preliminary diagnoses based on clinical symptoms, and a response was considered correct if the accurate diagnosis was included in the five generated. The model pairings were DeepSeek-R1-8B vs. Llama3.1-8B, DeepSeek-R1-14B vs. Qwen2.5-14B, DeepSeek-R1-32B vs. Qwen2.5-32B, DeepSeek-R1-70B vs. Llama3.3-70B, and DeepSeek-R1-671B vs. DeepSeek-V3. All reasoning models except DeepSeek-R1-671B were distilled versions. Diagnostic accuracy was assessed using McNemar’s test for discordant pairs, with a significance threshold of 0.01. The results showed that DeepSeek-R1-671B significantly outperformed DeepSeek-V3 (95.45% vs. 88.18%; p = 0.008), while DeepSeek-R1-8B underperformed relative to Llama3.1-8B (47.27% vs. 64.54%; p = 0.003). No significant differences were observed for the mid-sized models. Subgroup analyses based on incidence and clinical specialties further supported these conclusions. Qualitative analysis of the chain-of-thought outputs in incorrect cases revealed three universally prevalent error modes across distilled models: Reasoning drift, Red-Flag recognition failure, and diagnostic priority inversion. The study concludes that the DeepSeek-R1-671B shows potential for medical diagnosis, but distilled models do not exceed their base models. Based on simulated clinical cases, our results do not support deploying distilled models for text-based diagnostic tasks without further validation on real patient data.

Polystyrene nanoplastics (PS-NPs) exposure can induce placental toxicity, but the molecular mechanism of PS-NP-induced placental toxicity is poorly understood. We selected PS-NPs (80 nm) as a model of plastic particles and treated the human trophoblast cell line JEG3 (6.25, 12.5, 25, 50, and 100 mg/L) and pregnant mice with PS-NPs (15, 30 and 60 mg/kg·bw). These results found that PS-NPs regulated key ferroptosis-associated proteins (GPX4, ACSL4, and FTH1), indicating that PS-NPs induced ferroptosis. Here, we show for the first time that PS-NPs trigger placental ferroptosis by disrupting the p62-Keap1 interaction, thereby blocking Nrf2 degradation. Although m6A modification of the p62-Keap1-Nrf2 axis has been described in other models, its involvement in nanoplastic-induced placental injury remains unexplored. Here, we showed that PS-NPs induce ferroptosis by promoting METTL3 expression in JEG cells and placental tissues. Furthermore, knockdown of METTL3 inhibited PS-NP-induced ferroptosis by regulating Keap1 m6A, which inhibited the p62-Keap1-Nrf2 pathway in vitro. PS-NP-induced ferroptosis was attenuated by METTL3 knockdown via the inhibition of Keap1 m6A, which regulates the p62-Keap1-Nrf2 pathway. Collectively, these findings indicate thatMETTL3-mediated m6A methylation induced by PS-NPs may represent a previously unrecognized molecular basis for placental injury.

To explore the genotype-phenotype correlation in a Chinese family with structural brain abnormalities due to variant of the TUBB gene. A family undergoing prenatal diagnosis at Beijing Obstetrics and Gynecology Hospital in October 2024 was selected as the study subject. Clinical data were collected. Amniotic fluid sample was subjected to chromosomal copy number variation sequencing (CNV-seq). Trio whole-exome sequencing (Trio-WES) was carried out on the amniotic fluid and parental blood samples, and candidate variant was verified by Sanger sequencing. This study was approved by the Medical Ethics Committee of the hospital (Ethics No.: 2023-KY-076-01). Both prenatal ultrasound and fetal MRI showed deviation of brain midline, unilateral lateral ventriculomegaly, and bilateral gyral asymmetry. Trio-WES revealed that the fetus has harbored a maternally derived heterozygous missense variant of the TUBB gene [NM_178014.4: c.155A>G (p.N52S)]. Sanger sequencing confirmed that the woman and a previously terminated fetus both harbored the same variant. Both the proband and two fetuses exhibited similar neuroimaging abnormalities including midline deviation and asymmetrical gyri. Based on the guidelines from the American College of Medical Genetics and Genomics (ACMG), the variant was classified as likely pathogenic (PM2_Supporting+PS2_Moderate+PS3). The heterozygous c.155A>G (p.N52S) variant was the TUBB gene probably underlay the pathogenesis of the structural brain abnormalities in this family. Above findings have expanded the phenotypic spectrum associated with the variant and facilitated the prenatal diagnosis for this family.

To investigate the molecular epidemiology of children with phenylketonuria (PKU) in Gansu, China, providing foundational data for intervention strategies. A retrospective analysis was conducted on 1 159 PKU families who attended Gansu Provincial Maternity and Child Care Hospital from January 2012 to December 2024. Sanger sequencing, multiplex ligation-dependent probe amplification, whole exome sequencing, and deep intronic variant analysis were used to analyze the PAH gene. For the 1 159 children with PKU, 2 295 variants were identified in 2 318 alleles, resulting in a detection rate of 99.01%. The detection rates were 100% (914/914) in 457 classic PKU families, 99.45% (907/912) in 456 mild PKU families, and 96.34% (474/492) in 246 mild hyperphenylalaninemia families. The 2 295 variants detected comprised 208 distinct mutation types, among which c.728G>A (14.95%, 343/2 295) had the highest frequency, followed by c.611A>G (4.88%, 112/2 295) and c.721C>T (4.79%, 110/2 295). The cumulative frequency of the top 23 hotspot variants reached 70.28% (1 613/2 295), and most variant alleles were detected in exon 7 (29.19%, 670/2 295). Deep intronic variant analysis of the PAH gene can improve the genetic diagnostic rate of PKU. The development of targeted detection kits for PAH hotspot variants may enable precision screening programs and enhance preventive strategies for PKU.