Sleep and wake have global effects on brain physiology, from molecular changes and neuronal activities to synaptic plasticity. Sleep-wake homeostasis is maintained by the generation of a sleep need that accumulates during waking and dissipates during sleep. Here we investigate the molecular basis of sleep need using quantitative phosphoproteomic analysis of the sleep-deprived and Sleepy mouse models of increased sleep need. Sleep deprivation induces cumulative phosphorylation of the brain proteome, which dissipates during sleep. Sleepy mice, owing to a gain-of-function mutation in the Sik3 gene , have a constitutively high sleep need despite increased sleep amount. The brain proteome of these mice exhibits hyperphosphorylation, similar to that seen in the brain of sleep-deprived mice. Comparison of the two models identifies 80 mostly synaptic sleep-need-index phosphoproteins (SNIPPs), in which phosphorylation states closely parallel changes of sleep need. SLEEPY, the mutant SIK3 protein, preferentially associates with and phosphorylates SNIPPs. Inhibition of SIK3 activity reduces phosphorylation of SNIPPs and slow wave activity during non-rapid-eye-movement sleep, the best known measurable index of sleep need, in both Sleepy mice and sleep-deprived wild-type mice. Our results suggest that phosphorylation of SNIPPs accumulates and dissipates in relation to sleep need, and therefore SNIPP phosphorylation is a molecular signature of sleep need. Whereas waking encodes memories by potentiating synapses, sleep consolidates memories and restores synaptic homeostasis by globally downscaling excitatory synapses. Thus, the phosphorylation-dephosphorylation cycle of SNIPPs may represent a major regulatory mechanism that underlies both synaptic homeostasis and sleep-wake homeostasis.

No abstract

Machine learning (ML) is perhaps the most useful tool for the interpretation of large genomic datasets. However, the performance of a single machine learning method in genomic selection (GS) is currently unsatisfactory. To improve the genomic predictions, we constructed a stacking ensemble learning framework (SELF), integrating three machine learning methods, to predict genomic estimated breeding values (GEBVs). The present study evaluated the prediction ability of SELF by analyzing three real datasets, with different genetic architecture; comparing the prediction accuracy of SELF, base learners, genomic best linear unbiased prediction (GBLUP) and BayesB. For each trait, SELF performed better than base learners, which included support vector regression (SVR), kernel ridge regression (KRR) and elastic net (ENET). The prediction accuracy of SELF was, on average, 7.70% higher than GBLUP in three datasets. Except for the milk fat percentage (MFP) traits, of the German Holstein dairy cattle dataset, SELF was more robust than BayesB in all remaining traits. Therefore, we believed that SEFL has the potential to be promoted to estimate GEBVs in other animals and plants.

No abstract