NF1（Neurofibromin 1）和NF2（Neurofibromin 2）基因分别是1型和2型神经纤维瘤的致病基因，这两种疾病均属于常染色体显性遗传病。NF1和NF2基因在多种恶性肿瘤中也具有较高的突变频率，并与肿瘤的靶向治疗耐药密切相关。然而，NF1和NF2基因上存在大量的意义未明变异（Variant of Uncertain Significance，VUS），严重阻碍了临床诊断和治疗决策。本研究旨在通过开发优化的碱基编辑筛选技术和分析框架，系统性评估NF1和NF2基因变异的功能效应，为临床变异解读提供科学依据。

传统的碱基编辑高通量筛选技术存在多方面的局限性：PAM序列要求限制了可靶向的位点范围；sgRNA编辑效率的可变性导致筛选结果的假阴性率较高；编辑产物的复杂性增加了基因型-表型关联的难度；现有分析方法仅利用单一时间点的数据进行变异效应评估。针对这些挑战，我们进行了多项优化措施：首先，采用近乎无PAM序列限制的碱基编辑器，扩大了可研究的变异的范围；其次，在sgRNA表达载体中整合了编辑活性传感器系统和内部条形码，实现了编辑结果的直接检测和实验差异的有效控制；最后，开发了EDGE-BE算法，首次实现多时间序列数据与与sgRNA编辑活性整合分析，显著提高了变异效应的检测的准确性。

基于优化后的高通量碱基编辑筛选体系，我们对NF1基因上的15,876个变异和NF2基因上的7,386个变异进行了系统性功能评估，分别鉴定出505个NF1基因和166个NF2基因上的功能缺失（Loss of unction，LOF）变异。与非功能性变异相比，这些LOF变异在临床预测工具CADD和EVE评分中表现出较高分值，并在神经纤维瘤病患者和各种癌症队列中显著富集，尤其在黑色素瘤中富集程度很高。此外，我们分析了LOF变异的多种致病分子机制，包括转录调控异常、RNA剪接异常、蛋白功能活性受损以及蛋白二聚体界面稳定性破坏等。

利用单细胞转录组分析，我们进一步阐明了NF1功能变异在维罗非尼（Vemurafenib，PLX4032）耐药中的作用机制。在BRAF-V600E突变的黑色素瘤细胞中，NF1功能缺失变异能够克服PLX4032诱导的细胞周期G1期阻滞，并通过增强RAS活性重新激活丝裂原活化蛋白激酶（Mitogen-Activated Protein Kinase，MAPK）信号通路，最终导致耐药。此外，基于转录组偏移程度计算的E-distance评分与EDGE-BE评分相关性很高，验证了EDGE-BE方法的可靠性，同时证实了单细胞转录组技术在变异功能效应研究中的应用价值。

本研究构建了全面的NF1和NF2基因变异功能图谱，揭示了变异致病性的多种机制及其在药物耐药中的作用。我们开发的基于近乎无PAM序列限制的碱基编辑器的高通量筛选方法，结合EDGE-BE分析框架，为变异效应量化提供了一个通用的解决方案。筛选中获得的NF1和NF2基因变异功能性评分，不仅增进了对神经纤维瘤及相关癌症分子机制的理解，还为临床意义不明变异的解读提供了重要的工具和数据资源。本研究展示了碱基编辑筛选技术在解决疾病变异解读难题中的应用潜力，为其他疾病相关基因的功能变异研究提供了技术参考。

NF1 and NF2 genes are established causative genes for the autosomal dominant disorders neurofibromatosis type 1 and type 2, respectively. These genes exhibit high mutation frequencies across various malignancies and are closely associated with resistance to targeted therapies. However, the abundance of variants of uncertain significance (VUS) in NF1 and NF2 significantly impedes clinical diagnosis and therapeutic decision-making. This study aimed to systematically evaluate the functional effects of NF1 and NF2 variants through the development of advanced base editing screening technologies and analytical frameworks, thereby providing scientific evidence for clinical variant interpretation.

Traditional high-throughput base editing screening technologies present several limitations: PAM sequence requirements restrict the range of targetable sites; variability in sgRNA editing efficiency leads to high false-negative rates; complexity of editing products complicates genotype-phenotype associations; and analyses typically rely on single time-point data. To address these challenges, we implemented a series of technical innovations: first, we employed base editors with minimal PAM sequence restrictions, significantly expanding the targeting range; second, we integrated an editing activity sensor system and internal barcodes into the sgRNA expression vector, enabling direct detection of editing outcomes and effective control of experimental variations; finally, we developed the EDGE-BE algorithm, which for the first time integrates multi-time series data with sgRNA editing activity into a Bayesian framework, achieving precise quantification of variant effects.

Utilizing this optimized high-throughput base editing screening system, we systematically evaluated the functional impact of 15,876 variants in NF1 and 7,386 variants in NF2. We identified 505 loss-of-function (LOF) variants in NF1 and 166 LOF variants in NF2. These LOF variants demonstrated significant differences in clinical prediction tools such as CADD and EVE scores, and were significantly enriched in neurofibromatosis patients and various cancer cohorts, particularly in melanoma. Additionally, we revealed multiple molecular mechanisms underlying variant pathogenicity, including alterations in transcriptional control affecting gene expression, disruption of RNA splicing regulation leading to aberrant splicing, impact on protein functional activity, and disruption of protein dimer interface stability resulting in loss of protein function.

Single-cell transcriptome analysis further elucidated the mechanism by which NF1 functional variants contribute to vemurafenib (PLX4032) resistance. In BRAF-V600E mutated melanoma cells, NF1 loss-of-function variants overcame PLX4032-induced G1 cell cycle arrest by enhancing RAS activity and reactivating MAPK signaling, thereby mediating drug resistance. The E-distance score based on transcriptome shift calculations showed high concordance with EDGE-BE scores, validating the reliability of the EDGE-BE method while demonstrating the value of single-cell transcriptomics in characterizing variant functional effects.

This study established a comprehensive functional map of NF1 and NF2 gene variants, revealed multiple mechanisms of variant pathogenicity, and elucidated their role in drug resistance. Our optimized high-throughput screening method based on near-PAMless base editors and the EDGE-BE analytical framework provide a universal solution for variant effect quantification. This approach not only advances our understanding of the molecular mechanisms underlying neurofibromatosis and related cancers but also provides important tools and data resources for interpreting variants of uncertain significance, laying the foundation for individualized precision medicine. This study demonstrates the immense potential of base editing screening technologies in addressing the challenges of variant interpretation in complex genetic disorders and provides a technical paradigm for functional variant research in other disease-related genes.