当前大语言模型的兴起为自然语言处理、搜索引擎、生命科学研究等领域的研究者提供了新思路,但大语言模型存在资源消耗高、推理速度慢,难以在工业场景尤其是垂直领域应用等方面的缺点。针对这一问题,提出了一种多尺度卷积神经网络(convo...当前大语言模型的兴起为自然语言处理、搜索引擎、生命科学研究等领域的研究者提供了新思路,但大语言模型存在资源消耗高、推理速度慢,难以在工业场景尤其是垂直领域应用等方面的缺点。针对这一问题,提出了一种多尺度卷积神经网络(convolutional neural network,CNN)与双向长短期记忆神经网络(long short term memory,LSTM)融合的唐卡问句分类模型,本文模型将数据的全局特征与局部特征进行融合实现唐卡问句分类任务,全局特征反映数据的本质特点,局部特征关注数据中易被忽视的部分,将二者以拼接的方式融合以丰富句子的特征表示。通过在Thangka数据集与THUCNews数据集上进行实验,结果表明,本文模型相较于Bert模型在精确度上略优,在训练时间上缩短了1/20,运算推理时间缩短了1/3。在公开数据集上的实验表明,本文模型在文本分类任务上也表现出了较好的适用性和有效性。展开更多
Background Congenital cataract is a highly heterogeneous disorder at both the genetic and phenotypic levels. This study was conducted to identify disease locus for autosomal dominant congenital cataracts in a four gen...Background Congenital cataract is a highly heterogeneous disorder at both the genetic and phenotypic levels. This study was conducted to identify disease locus for autosomal dominant congenital cataracts in a four generation Chinese family. Methods Family history and clinical data were recorded. All the members were genotyped with microsatellite markers which are close to the known genetic loci for autosomal congenital cataracts. Two-point Lod scores were obtained using the MLINK of the LINKAGE program package （vet 5.1）. Candidate genes were amplified by polymerase chain reaction （PCR） and direct cycle sequencing.Results The maximum Lod score of Zmax=2.11 was obtained with three microsatellite markers D22S258, D22S315, and D22S1163 at recombination fraction θ= 0. Haplotype analysis showed that the disease gene was localized to a 18.5 Mbp region on chromosome 22 flanked by markers D22S1174 and D22S270, spanning the β-crystallin gene cluster. A c.752T→C mutation in exon 6 of CRYBB1 gene, which resulted in a heterozygous S228P mutation in predicted protein, was found to cosegregate with cataract in the family.Conclusions This study identified a novel mutation in CRYBB1 gene in a Chinese family with autosomal dominant congenital cataract. These results provide strong evidence that CRYBB1 is a pathogenic gene for congenital cataract.展开更多
Background Activation of c-Jun NH2-terminal kinase （JNK） has been implicated in neuron apoptosis as well as autophagy in response to various stressors after traumatic brain injury （TBI）. However, the underlying mo...Background Activation of c-Jun NH2-terminal kinase （JNK） has been implicated in neuron apoptosis as well as autophagy in response to various stressors after traumatic brain injury （TBI）. However, the underlying molecular pathway remains unclear. Our study assessed whether JNK-mediated p53 phosphorylation might be an important mechanism for enhancing neuron autophagy in response to TBI. Methods A total of 186 male Sprague-Dawley （SD） rats （300-350 g） were used in this study. By randomized block method rats were randomly divided into four groups： sham-operated （n=46）, TBI （n=60）, TBI ＋ dimethyl sulfoxide （DMSO） （n=40）, and TBI ＋ SP600125 （n=40）. JNK was treated with SP600125, a specific JNK inhibitor. JNK, p-P53, Beclin-1, damage-regulated autophagy modulator （DRAM） and p-bcl-2 were evaluated by Western blotting analysis. The cellular localization and expression of Beclin-1 and DRAM was observed by immunofluorescence and immunohistochemistry, and the expression of Beclin-l-Bcl-2/Bcl-xL complexes was evaluated by immunoprecipitation. Multiple-group comparisons were conducted using analysis of variance （ANOVA）. P values of less than 0.05 were considered statistically significant. Results It was observed that the expression of JNK, p-P53, Beclin-1, DRAM and p-bcl-2 was increasing after TBI, and the expression of Beclin-1 and DRAM was mainly located in the cytoplasm of neurons. But these were significantly inhibited in SP600125 group compared with sham group and TBI＋SP600125 group （P 〈0.05）. The expression of Beclin-l-Bcl-2/Bcl-xL complexes was reduced after TBI. Conclusion JNK-mediated p53 phosphorylation might be an important mechanism for enhancing neuron autophagy in response to TBI.展开更多
文摘当前大语言模型的兴起为自然语言处理、搜索引擎、生命科学研究等领域的研究者提供了新思路,但大语言模型存在资源消耗高、推理速度慢,难以在工业场景尤其是垂直领域应用等方面的缺点。针对这一问题,提出了一种多尺度卷积神经网络(convolutional neural network,CNN)与双向长短期记忆神经网络(long short term memory,LSTM)融合的唐卡问句分类模型,本文模型将数据的全局特征与局部特征进行融合实现唐卡问句分类任务,全局特征反映数据的本质特点,局部特征关注数据中易被忽视的部分,将二者以拼接的方式融合以丰富句子的特征表示。通过在Thangka数据集与THUCNews数据集上进行实验,结果表明,本文模型相较于Bert模型在精确度上略优,在训练时间上缩短了1/20,运算推理时间缩短了1/3。在公开数据集上的实验表明,本文模型在文本分类任务上也表现出了较好的适用性和有效性。
基金This study was supported by grants from National Natural Science Foundation(No.30471864)National Key Program(No.2003DEA3N026)of China.
文摘Background Congenital cataract is a highly heterogeneous disorder at both the genetic and phenotypic levels. This study was conducted to identify disease locus for autosomal dominant congenital cataracts in a four generation Chinese family. Methods Family history and clinical data were recorded. All the members were genotyped with microsatellite markers which are close to the known genetic loci for autosomal congenital cataracts. Two-point Lod scores were obtained using the MLINK of the LINKAGE program package （vet 5.1）. Candidate genes were amplified by polymerase chain reaction （PCR） and direct cycle sequencing.Results The maximum Lod score of Zmax=2.11 was obtained with three microsatellite markers D22S258, D22S315, and D22S1163 at recombination fraction θ= 0. Haplotype analysis showed that the disease gene was localized to a 18.5 Mbp region on chromosome 22 flanked by markers D22S1174 and D22S270, spanning the β-crystallin gene cluster. A c.752T→C mutation in exon 6 of CRYBB1 gene, which resulted in a heterozygous S228P mutation in predicted protein, was found to cosegregate with cataract in the family.Conclusions This study identified a novel mutation in CRYBB1 gene in a Chinese family with autosomal dominant congenital cataract. These results provide strong evidence that CRYBB1 is a pathogenic gene for congenital cataract.
基金This work was supported by a grant from Hebei Province Natural Science Foundation （No. C2009001247）. The authors declare that there are no conflicts of interest.Acknowledgments： We thank Dr. CUI Jian-zhong and Dr. GAO Jun-ling for their help and generous gift of antibodies for Western blotting.
文摘Background Activation of c-Jun NH2-terminal kinase （JNK） has been implicated in neuron apoptosis as well as autophagy in response to various stressors after traumatic brain injury （TBI）. However, the underlying molecular pathway remains unclear. Our study assessed whether JNK-mediated p53 phosphorylation might be an important mechanism for enhancing neuron autophagy in response to TBI. Methods A total of 186 male Sprague-Dawley （SD） rats （300-350 g） were used in this study. By randomized block method rats were randomly divided into four groups： sham-operated （n=46）, TBI （n=60）, TBI ＋ dimethyl sulfoxide （DMSO） （n=40）, and TBI ＋ SP600125 （n=40）. JNK was treated with SP600125, a specific JNK inhibitor. JNK, p-P53, Beclin-1, damage-regulated autophagy modulator （DRAM） and p-bcl-2 were evaluated by Western blotting analysis. The cellular localization and expression of Beclin-1 and DRAM was observed by immunofluorescence and immunohistochemistry, and the expression of Beclin-l-Bcl-2/Bcl-xL complexes was evaluated by immunoprecipitation. Multiple-group comparisons were conducted using analysis of variance （ANOVA）. P values of less than 0.05 were considered statistically significant. Results It was observed that the expression of JNK, p-P53, Beclin-1, DRAM and p-bcl-2 was increasing after TBI, and the expression of Beclin-1 and DRAM was mainly located in the cytoplasm of neurons. But these were significantly inhibited in SP600125 group compared with sham group and TBI＋SP600125 group （P 〈0.05）. The expression of Beclin-l-Bcl-2/Bcl-xL complexes was reduced after TBI. Conclusion JNK-mediated p53 phosphorylation might be an important mechanism for enhancing neuron autophagy in response to TBI.
