Protein aggregates,mitochondrial import stress and neurodegenerative disorders:A salient hallmark of several neurodegenerative diseases,including Parkinson’s disease,is the abundance of protein aggregates(Goiran et a...Protein aggregates,mitochondrial import stress and neurodegenerative disorders:A salient hallmark of several neurodegenerative diseases,including Parkinson’s disease,is the abundance of protein aggregates(Goiran et al.,2022).This molecular event is believed to lead to activation of stress pathways ultimately resulting in cellular dysfunction(Eldeeb et al.,2022).Accordingly,many lines of research investigations focused on dampening the formation of protein aggregates or augmenting the clearance of protein aggregates as a potential therapeutic strategy to counteract the progression of neurodegenerative diseases,albeit with little success(Costa-Mattioli and Walter,2020).Cell stress cues such as the accumulation of protein aggregates lead to the activation of stress response pathways that aid cells in responding to the damage.Despite the notion that the transient activation of these pathways helps cells cope with stressors,persistent activation can induce unwanted apoptosis of cells and reduce overall tissue strength as well as lead to an accumulation of aggregation-prone proteins(Hetz and Papa,2018).Mutations in proteins involved in stress signaling termination can cause conditions like ataxia and early-onset dementia(Conroy et al.,2014).Therefore,it is crucial for stress response signaling to be turned off once conditions have improved.Nevertheless,the mechanisms by which cells silence these signals are still elusive.展开更多
文摘Protein aggregates,mitochondrial import stress and neurodegenerative disorders:A salient hallmark of several neurodegenerative diseases,including Parkinson’s disease,is the abundance of protein aggregates(Goiran et al.,2022).This molecular event is believed to lead to activation of stress pathways ultimately resulting in cellular dysfunction(Eldeeb et al.,2022).Accordingly,many lines of research investigations focused on dampening the formation of protein aggregates or augmenting the clearance of protein aggregates as a potential therapeutic strategy to counteract the progression of neurodegenerative diseases,albeit with little success(Costa-Mattioli and Walter,2020).Cell stress cues such as the accumulation of protein aggregates lead to the activation of stress response pathways that aid cells in responding to the damage.Despite the notion that the transient activation of these pathways helps cells cope with stressors,persistent activation can induce unwanted apoptosis of cells and reduce overall tissue strength as well as lead to an accumulation of aggregation-prone proteins(Hetz and Papa,2018).Mutations in proteins involved in stress signaling termination can cause conditions like ataxia and early-onset dementia(Conroy et al.,2014).Therefore,it is crucial for stress response signaling to be turned off once conditions have improved.Nevertheless,the mechanisms by which cells silence these signals are still elusive.
文摘目的针对联邦学习中多中心医学数据的异质性特征导致全局模型性能不佳的问题,提出一种基于特征迁移的自适应个性化联邦学习算法(adaptive personalized federated learning via feature transfer,APFFT)。方法首先,为降低全局模型中异质性特征信息影响,提出鲁棒特征选择网络(robust feature selection network,RFS-Net)构建个性化本地模型。RFS-Net通过学习两个迁移权重分别确定全局模型向本地模型迁移时的有效特征以及特征迁移的目的地,并构建基于迁移权重的迁移损失函数以加强本地模型对全局模型中有效特征的注意力,从而构建个性化本地模型。然后,为过滤各本地模型中异质性特征信息,利用自适应聚合网络(adaptive aggregation network,AANet)聚合全局模型。AA-Net基于全局模型交叉熵变化更新迁移权重并构建聚合损失,使各本地模型向全局模型迁移鲁棒特征,提高全局模型的特征表达能力。结果在3种医学图像分类任务上与4种现有方法进行比较实验,在肺结核肺腺癌分类任务中,各中心曲线下面积(area under the curve,AUC)分别为0.7915,0.7981,0.7600,0.7057和0.8069;在乳腺癌组织学图像分类任务中,各中心准确率分别为0.9849、0.9808、0.9835、0.9826和0.9834;在肺结节良恶性分类任务中,各中心AUC分别为0.8097,0.8498,0.7848和0.7923。结论所提出的联邦学习方法,降低了多中心的异质性特征影响,实现基于鲁棒特征的个性化本地模型自适应构建和全局模型自适应聚合,模型性能有较大提升。
