Prediction of Sleep Stages Via Deep Learning Using Smartphone Audio Recordings in Home Environments: Model Development and Validation.

BACKGROUND: The growing public interest and awareness regarding the significance of sleep is driving the demand for sleep monitoring at home. In addition to various commercially available wearable and nearable devices, sound-based sleep staging via deep learning is emerging as a decent alternative for their convenience and potential accuracy. However, sound-based sleep staging has only been studied using in-laboratory sound data. In real-world sleep environments (homes), there is abundant background noise, in contrast to quiet, controlled environments such as laboratories. The use of sound-based sleep staging at homes has not been investigated while it is essential for practical use on a daily basis. Challenges are the lack of and the expected huge expense of acquiring a sufficient size of home data annotated with sleep stages to train a large-scale neural network. OBJECTIVE: This study aims to develop and validate a deep learning method to perform sound-based sleep staging using audio recordings achieved from various uncontrolled home environments. METHODS: To overcome the limitation of lacking home data with known sleep stages, we adopted advanced training techniques and combined home data with hospital data. The training of the model consisted of 3 components: (1) the original supervised learning using 812 pairs of hospital polysomnography (PSG) and audio recordings, and the 2 newly adopted components; (2) transfer learning from hospital to home sounds by adding 829 smartphone audio recordings at home; and (3) consistency training using augmented hospital sound data. Augmented data were created by adding 8255 home noise data to hospital audio recordings. Besides, an independent test set was built by collecting 45 pairs of overnight PSG and smartphone audio recording at homes to examine the performance of the trained model. RESULTS: The accuracy of the model was 76.2% (63.4% for wake, 64.9% for rapid-eye movement [REM], and 83.6% for non-REM) for our test set. The macro F1-score and mean per-class sensitivity were 0.714 and 0.706, respectively. The performance was robust across demographic groups such as age, gender, BMI, or sleep apnea severity (accuracy 73.4%-79.4%). In the ablation study, we evaluated the contribution of each component. While the supervised learning alone achieved accuracy of 69.2% on home sound data, adding consistency training to the supervised learning helped increase the accuracy to a larger degree (+4.3%) than adding transfer learning (+0.1%). The best performance was shown when both transfer learning and consistency training were adopted (+7.0%). CONCLUSIONS: This study shows that sound-based sleep staging is feasible for home use. By adopting 2 advanced techniques (transfer learning and consistency training) the deep learning model robustly predicts sleep stages using sounds recorded at various uncontrolled home environments, without using any special equipment but smartphones only.

Oriented insertion of phi29 N-hexahistidine-tagged gp10 connector protein assemblies into C20BAS bolalipid membrane vesicles.

Ni(2+):NTA-PEG600-grafted glass surfaces are capable of immobilizing N-his6 gp10 connector protein assemblies from the phi29 DNA packaging motor and mediating their transplantation into bolalipid vesicles whose membrane thickness is compatible with the hydrophobic domain of the gp10 assemblies.

Cytoplasmic delivery of liposomal contents mediated by an acid-labile cholesterol-vinyl ether-PEG conjugate.

An acid-cleavable PEG lipid, 1'-(4'-cholesteryloxy-3'-butenyl)-omega-methoxy-polyethylene[112] glycolate (CVEP), has been developed that produces stable liposomes when dispersed as a minor component (0.5-5 mol %) in 1,2-dioleoyl-sn-glycero-3-phosphoethanolamine (DOPE). Cleavage of CVEP at mildly acidic pHs results in dePEGylation of the latently fusogenic DOPE liposomes, thereby triggering the onset of content release. This paper describes the synthesis of CVEP via a six-step sequence starting from the readily available precursors 1,4-butanediol, cholesterol, and mPEG acid. The hydrolysis rates and release kinetics from CVEP/DOPE liposome dispersions as a function of CVEP loading, as well as the cryogenic transmission electron microscopy and pH-dependent monolayer properties of 9:91 CVEP/DOPE mixtures, also are reported. When folate receptor-positive KB cells were exposed to calcein-loaded 5:95 CVEP/DOPE liposomes containing 0.1 mol % folate-modified 1,2-distearoyl-sn-glycero-3-phosphoethanolamine-N-polyethylene[76] glycolamide (folate-PEG-DSPE), delivery of the calcein cargo to the cytoplasm of the cells was observed as determined by fluorescence microscopy and flow cytometry. Fluorescence resonance energy transfer analysis of lipid mixing in these cells was consistent with membrane-membrane fusion between the liposome and endosomal membranes.

Synthesis and self-assembly of glycal-based bolaforms.

Glycal-based bolaforms serve as synthetically flexible components of molecular self-assembly. The compounds are prepared in good yield by a Ferrier reaction between triacetylglucal or -galactal or diacetylxylal and a long chain alpha,omega-diol, followed by deacetylation under Zemplen conditions. The reactions are stereoselective and preferentially afford the alpha-diastereomer. The bolaforms undergo self-assembly in water or water/dioxane solution to give a variety of nanostructures. In solution, bolaforms with C8 or C10 chains between glucal headgroups form nanoscale vesicles. In contrast, bolaforms with C12 chains exhibit lower solubility and a dynamic self-assembly, forming several different nanoscale structures. However, the solid-state structures of C12 bolaform isomers adopt shapes very similar to those of bolaforms possessing more extensive hydrogen-bonding networks, indicating that multiple hydrogen bonds in solution are important to formation of stable, discrete nanostructures but that only a few key intermolecular interactions between bolaform headgroups are necessary to determine the structure in the solid state. The diversity and differentiation of the functional groups present in glycal-based bolaforms suggest that they could be useful probes of the various noncovalent forces controlling the structure of new nanomaterials.

Specific adsorption of histidine-tagged proteins on silica surfaces modified with Ni2+/NTA-derivatized poly(ethylene glycol).

Silica surfaces modified with nitrilotriacetic acid (NTA)-polyethylene glycol (PEG) derivatives were used to immobilize hexahistidine-tagged green fluorescent protein (His6-GFP), biotin/streptavidin-AlexaFluor555 (His6-biotin/SA-AF), and gramicidin A-containing vesicles (His6-gA). Three types of surface-reactive PEG derivatives-NTA-PEG3400-Si(OMe)3, NTA-PEG3400-vinylsulfone, and mPEG5000-Si(OMe)3 (control)-were grafted onto silica and tested for their ability to capture His6-tag species via His6/Ni2+/NTA chelation. The composition and thicknesses of the PEG-modified surfaces were characterized using X-ray photoelectron spectroscopy, contact angle, and ellipsometry. Protein capture efficiencies of the NTA-PEG-grafted surfaces were evaluated by measuring fluorescence intensities of these surfaces after exposure to His6-tag species. XPS and ellipsometry data indicate that surface adsorption occurs via specific interactions between the His6-tag and the Ni2+/NTA-PEG-grafted surface. Protein immobilization was most effective for NTA-PEG3400-Si(OMe)3-modified surfaces, with maximal areal densities achieved at 45 pmol/cm2 for His6-GFP and 95 fmol/cm2 for His6-biotin/SA-AF. Lipid vesicles containing His6-gA in a 1:375 gA/lipid ratio could also be immobilized on Ni2+/NTA-PEG3400-Si(OMe)3-modified surfaces at 0.5 mM total lipid. Our results suggest that NTA-PEG-Si(OMe)3 conjugates may be useful tools for immobilizing His6-tag proteins on solid surfaces to produce protein-functionalized surfaces.

Interactions between pH-sensitive liposomes and model membranes.

The structure and dynamics of two different pH-sensitive liposome systems were investigated by means of cryo-transmission electron microscopy and different photophysical techniques. Both systems consisted of dioleoylphosphatidylethanolamine (DOPE) and contained either oleic acid (OA) or a novel acid-labile polyethylene glycol-conjugated lipid (DHCho-MPEG5000) as stabiliser. Proton induced leakage, lipid mixing and structural changes were studied in the absence and presence of EPC liposomes, as well as in the presence of liposomes designed to model the endosome membrane. Neither DHCho-MPEG5000- nor OA-stabilised liposomes showed any tendency for fusion with pure EPC liposomes or endosome-like liposomes composed of EPC/DOPE/SM/Cho (40/20/6/34 mol.%). Our investigations showed, however, that incorporation of lipids from the pH-sensitive liposomes into the endosome membrane may lead to increased permeability and formation of non-lamellar structures. Taken together the results suggest that the observed ability of DOPE-containing liposomes to mediate cytoplasmic delivery of hydrophilic molecules cannot be explained by a mechanism based on a direct, and non-leaky, fusion between the liposome and endosome membranes. A mechanism involving destabilisation of the endosome membrane due to incorporation of DOPE, seems more plausible.