Single-sequence protein structure prediction by integrating protein language models.

Protein structure prediction has been greatly improved by deep learning in the past few years. However, the most successful methods rely on multiple sequence alignment (MSA) of the sequence homologs of the protein under prediction. In nature, a protein folds in the absence of its sequence homologs and thus, a MSA-free structure prediction method is desired. Here, we develop a single-sequence-based protein structure prediction method RaptorX-Single by integrating several protein language models and a structure generation module and then study its advantage over MSA-based methods. Our experimental results indicate that in addition to running much faster than MSA-based methods such as AlphaFold2, RaptorX-Single outperforms AlphaFold2 and other MSA-free methods in predicting the structure of antibodies (after fine-tuning on antibody data), proteins of very few sequence homologs, and single mutation effects. By comparing different protein language models, our results show that not only the scale but also the training data of protein language models will impact the performance. RaptorX-Single also compares favorably to MSA-based AlphaFold2 when the protein under prediction has a large number of sequence homologs.

Sustained AhR activity programs memory fate of early effector CD8+ T cells.

Identification of mechanisms that program early effector T cells to either terminal effector T (Teff) or memory T (Tm) cells has important implications for protective immunity against infections and cancers. Here, we show that the cytosolic transcription factor aryl hydrocarbon receptor (AhR) is used by early Teff cells to program memory fate. Upon antigen engagement, AhR is rapidly up-regulated via reactive oxygen species signaling in early CD8+ Teff cells, which does not affect the effector response, but is required for memory formation. Mechanistically, activated CD8+ T cells up-regulate HIF-1α to compete with AhR for HIF-1ß, leading to the loss of AhR activity in HIF-1αhigh short-lived effector cells, but sustained in HIF-1αlow memory precursor effector cells (MPECs) with the help of autocrine IL-2. AhR then licenses CD8+ MPECs in a quiescent state for memory formation. These findings partially resolve the long-standing issue of how Teff cells are regulated to differentiate into memory cells.

Anterior cingulate cortex orexin signaling mediates early-life stress-induced social impairment in females.

Early-life stress has long-term impacts on the structure and function of the anterior cingulate cortex (ACC), and raises the risk of adult neuropsychiatric disorders including social dysfunction. The underlying neural mechanisms, however, are still uncertain. Here, we show that, in female mice, maternal separation (MS) during the first three postnatal weeks results in social impairment accompanied with hypoactivity in pyramidal neurons (PNs) of the ACC. Activation of ACC PNs ameliorates MS-induced social impairment. Neuropeptide Hcrt, which encodes hypocretin (orexin), is the top down-regulated gene in the ACC of MS females. Activating ACC orexin terminals enhances the activity of ACC PNs and rescues the diminished sociability observed in MS females via an orexin receptor 2 (OxR2)-dependent mechanism. Our results suggest orexin signaling in the ACC is critical in mediating early-life stress-induced social impairment in females.

Application of improved GalNAc conjugation in development of cost-effective siRNA therapies targeting cardiovascular diseases.

N-Acetylgalactosamine (GalNAc)-conjugated small interfering RNA (siRNA) therapies have received approval for treating both orphan and prevalent diseases. To improve in vivo efficacy and streamline the chemical synthesis process for efficient and cost-effective manufacturing, we conducted this study to identify better designs of GalNAc-siRNA conjugates for therapeutic development. Here, we present data on redesigned GalNAc-based ligands conjugated with siRNAs against angiopoietin-like 3 (ANGPTL3) and lipoprotein (a) (Lp(a)), two target molecules with the potential to address large unmet medical needs in atherosclerotic cardiovascular diseases. By attaching a novel pyran-derived scaffold to serial monovalent GalNAc units before solid-phase oligonucleotide synthesis, we achieved increased GalNAc-siRNA production efficiency with fewer synthesis steps compared to the standard triantennary GalNAc construct L96. The improved GalNAc-siRNA conjugates demonstrated equivalent or superior in vivo efficacy compared to triantennary GalNAc-conjugated siRNAs.

Global airborne bacterial community-interactions with Earth's microbiomes and anthropogenic activities.

Airborne bacteria are an influential component of the Earth's microbiomes, but their community structure and biogeographic distribution patterns have yet to be understood. We analyzed the bacterial communities of 370 air particulate samples collected from 63 sites around the world and constructed an airborne bacterial reference catalog with more than 27 million nonredundant 16S ribosomal RNA (rRNA) gene sequences. We present their biogeographic pattern and decipher the interlacing of the microbiome co-occurrence network with surface environments of the Earth. While the total abundance of global airborne bacteria in the troposphere (1.72 × 1024 cells) is 1 to 3 orders of magnitude lower than that of other habitats, the number of bacterial taxa (i.e., richness) in the atmosphere (4.71 × 108 to 3.08 × 109) is comparable to that in the hydrosphere, and its maximum occurs in midlatitude regions, as is also observed in other ecosystems. The airborne bacterial community harbors a unique set of dominant taxa (24 species); however, its structure appears to be more easily perturbed, due to the more prominent role of stochastic processes in shaping community assembly. This is corroborated by the major contribution of surface microbiomes to airborne bacteria (averaging 46.3%), while atmospheric conditions such as meteorological factors and air quality also play a role. Particularly in urban areas, human impacts weaken the relative importance of plant sources of airborne bacteria and elevate the occurrence of potential pathogens from anthropogenic sources. These findings serve as a key reference for predicting planetary microbiome responses and the health impacts of inhalable microbiomes with future changes in the environment.

Hybrid Ghost Phonon Polaritons in Thin-Film Heterostructure.

Ghost phonon polaritons (g-PhPs), a unique class of phonon polaritons in the infrared, feature ultralong diffractionless propagation (>20 µm) across the surface and tilted wavefronts in the bulk. Here, we study hybrid g-PhPs in a heterostructure of calcite and an ultrathin film of the phase change material (PCM) In3SbTe2, where the optical field is bound in the PCM film with enhanced confinement compared with conventional g-PhPs. Near-field optical images for hybrid g-PhPs reveal a lemniscate pattern in the momentum distribution. We fabricated In3SbTe2 gratings and investigated how different orientations and periodicities of gratings impact the propagation of hybrid g-PhPs. As the grating period decreases to zero, the wavefront of hybrid g-PhPs can be dynamically steered by varying the grating orientation. Our results highlight the promise of hybrid g-PhPs with tunable functionalities for nanophotonic studies.

Observation of Near-Field Thermal Radiation between Coplanar Nanodevices with Subwavelength Dimensions.

With the continuous advancement of nanotechnology, nanodevices have become crucial components in computing, sensing, and energy conversion applications. The structures of nanodevices typically possess subwavelength dimensions and separations, which pose significant challenges for understanding energy transport phenomena in nanodevices. Here, on the basis of a judiciously designed thermal photonic nanodevice, we report the first measurement of near-field energy transport between two coplanar subwavelength structures over temperature bias up to ∼190 K. Our experimental results demonstrate a 20-fold enhancement in energy transfer beyond blackbody radiation. In contrast with the well-established near-field interactions between two semi-infinite bodies, the subwavelength confinements in nanodevices lead to increased polariton scattering and reduction of supporting photonic modes and, therefore, a lower energy flow at a given separation. Our work unveils exciting opportunities for the rational design of nanodevices, particularly for coplanar near-field energy transport, with important implications for the development of efficient nanodevices for energy harvesting and thermal management.

Zinc finger BED-type containing 3 promotes hepatic steatosis by interacting with polypyrimidine tract-binding protein 1.

AIMS/HYPOTHESIS: The relationship between metabolic dysfunction-associated steatotic liver disease (MASLD) and type 2 diabetes mellitus, insulin resistance and the metabolic syndrome is well established. While zinc finger BED-type containing 3 (ZBED3) has been linked to type 2 diabetes mellitus and the metabolic syndrome, its role in MASLD remains unclear. In this study, we aimed to investigate the function of ZBED3 in the context of MASLD. METHODS: Expression levels of ZBED3 were assessed in individuals with MASLD, as well as in cellular and animal models of MASLD. In vitro and in vivo analyses were conducted using a cellular model of MASLD induced by NEFA and an animal model of MASLD induced by a high-fat diet (HFD), respectively, to investigate the role of ZBED3 in MASLD. ZBED3 expression was increased by lentiviral infection or tail-vein injection of adeno-associated virus. RNA-seq and bioinformatics analysis were employed to examine the pathways through which ZBED3 modulates lipid accumulation. Findings from these next-generation transcriptome sequencing studies indicated that ZBED3 controls SREBP1c (also known as SREBF1; a gene involved in fatty acid de novo synthesis); thus, co-immunoprecipitation and LC-MS/MS were utilised to investigate the molecular mechanisms by which ZBED3 regulates the sterol regulatory element binding protein 1c (SREBP1c). RESULTS: In this study, we found that ZBED3 was significantly upregulated in the liver of individuals with MASLD and in MASLD animal models. ZBED3 overexpression promoted NEFA-induced triglyceride accumulation in hepatocytes in vitro. Furthermore, the hepatocyte-specific overexpression of Zbed3 promoted hepatic steatosis. Conversely, the hepatocyte-specific knockout of Zbed3 resulted in resistance of HFD-induced hepatic steatosis. Mechanistically, ZBED3 interacts directly with polypyrimidine tract-binding protein 1 (PTBP1) and affects its binding to the SREBP1c mRNA precursor to regulate SREBP1c mRNA stability and alternative splicing. CONCLUSIONS/INTERPRETATION: This study indicates that ZBED3 promotes hepatic steatosis and serves as a critical regulator of the progression of MASLD. DATA AVAILABILITY: RNA-seq data have been deposited in the NCBI Gene Expression Omnibus ( www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE231875 ). MS proteomics data have been deposited to the ProteomeXchange Consortium via the iProX partner repository ( https://proteomecentral.proteomexchange.org/cgi/GetDataset?ID=PXD041743 ).

Singlet oxygen induces cell wall thickening and stomatal density reducing by transcriptome reprogramming.

Singlet oxygen (1O2) has a very short half-life of 10-5 s; however, it is a strong oxidant that causes growth arrest and necrotic lesions on plants. Its signaling pathway remains largely unknown. The Arabidopsis flu (fluorescent) mutant accumulates a high level of 1O2 and shows drastic changes in nuclear gene expression. Only two plastid proteins, EX1 (executer 1) and EX2 (executer 2), have been identified in the singlet oxygen signaling. Here, we found that the transcription factor abscisic acid insensitive 4 (ABI4) binds the promoters of genes responsive to 1O2-signals. Inactivation of the ABI4 protein in the flu/abi4 double mutant was sufficient to compromise the changes of almost all 1O2-responsive-genes and rescued the lethal phenotype of flu grown under light/dark cycles, similar to the flu/ex1/ex2 triple mutant. In addition to cell death, we reported for the first time that 1O2 also induces cell wall thickening and stomatal development defect. Contrastingly, no apparent growth arrest was observed for the flu mutant under normal light/dim light cycles, but the cell wall thickening (doubled) and stomatal density reduction (by two-thirds) still occurred. These results offer a new idea for breeding stress tolerant plants.

Development and validation of a perivascular space segmentation method in multi-center datasets.

BACKGROUND: Perivascular spaces (PVS) visible on magnetic resonance imaging (MRI) are significant markers associated with various neurological diseases. Although quantitative analysis of PVS may enhance sensitivity and improve consistency across studies, the field lacks a universally validated method for analyzing images from multi-center studies. METHODS: We annotated PVS on multi-center 3D T1-weighted (T1w) images acquired using scanners from three major vendors (Siemens, General Electric, and Philips). A neural network, mcPVS-Net (multi-center PVS segmentation network), was trained using data from 40 subjects and then tested in a separate cohort of 15 subjects. We assessed segmentation accuracy against ground truth masks tailored for each scanner vendor. Additionally, we evaluated the agreement between segmented PVS volumes and visual scores for each scanner. We also explored correlations between PVS volumes and various clinical factors such as age, hypertension, and white matter hyperintensities (WMH) in a larger sample of 1020 subjects. Furthermore, mcPVS-Net was applied to a new dataset comprising both T1w and T2-weighted (T2w) images from a United Imaging scanner to investigate if PVS volumes could discriminate between subjects with differing visual scores. We also compared the mcPVS-Net with a previously published method that segments PVS from T1 images. RESULTS: In the test dataset, mcPVS-Net achieved a mean DICE coefficient of 0.80, with an average Precision of 0.81 and Recall of 0.79, indicating good specificity and sensitivity. The segmented PVS volumes were significantly associated with visual scores in both the basal ganglia (r = 0.541, p < 0.001) and white matter regions (r = 0.706, p < 0.001), and PVS volumes were significantly different among subjects with varying visual scores. Segmentation performance was consistent across different scanner vendors. PVS volumes exhibited significant associations with age, hypertension, and WMH. In the United Imaging scanner dataset, PVS volumes showed good associations with PVS visual scores evaluated on either T1w or T2w images. Compared to a previously published method, mcPVS-Net showed a higher accuracy and improved PVS segmentation in the basal ganglia region. CONCLUSION: The mcPVS-Net demonstrated good accuracy for segmenting PVS from 3D T1w images. It may serve as a useful tool for future PVS research.

Higher intracranial arterial pulsatility is associated with presumed imaging markers of the glymphatic system: An explorative study.

BACKGROUND: Arterial pulsation has been suggested as a key driver of paravascular cerebrospinal fluid flow, which is the foundation of glymphatic clearance. However, whether intracranial arterial pulsatility is associated with glymphatic markers in humans has not yet been studied. METHODS: Seventy-three community participants were enrolled in the study. 4D phase-contrast magnetic resonance imaging (MRI) was used to quantify the hemodynamic parameters including flow pulsatility index (PIflow) and area pulsatility index (PIarea) from 13 major intracerebral arterial segments. Three presumed neuroimaging markers of the glymphatic system were measured: including dilation of perivascular space (PVS), diffusivity along the perivascular space (ALPS), and volume fraction of free water (FW) in white matter. We explored the relationships between PIarea, PIflow, and the presumed glymphatic markers, controlling for related covariates. RESULTS: PIflow in the internal carotid artery (ICA) C2 segment (OR, 1.05; 95 % CI, 1.01-1.10, per 0.01 increase in PI) and C4 segment (OR, 1.05; 95 % CI, 1.01-1.09) was positively associated with the dilation of basal ganglia PVS, and PIflow in the ICA C4 segment (OR, 1.06, 95 % CI, 1.02-1.10) was correlated with the dilation of PVS in the white matter. ALPS was associated with PIflow in the basilar artery (ß, -0.273, p, 0.046) and PIarea in the ICA C2 (ß, -0.239, p, 0.041) and C7 segments (ß, -0.238, p, 0.037). CONCLUSIONS: Intracranial arterial pulsatility was associated with presumed neuroimaging markers of the glymphatic system, but the results were not consistent across different markers. Further studies are warranted to confirm these findings.

Reticular Ratchets for Directing Electrochemiluminescence.

Electrochemiluminescence (ECL) involves charge transfer between electrochemical redox intermediates to produce an excited state for light emission. Ensuring precise control of charge transfer is essential for decoding ECL fundamentals, yet guidelines on how to achieve this for conventional emitters remain unexplored. Molecular ratchets offer a potential solution, as they enable the directional transfer of energy or chemicals while impeding the reverse movement. Herein, we designed 10 pairs of imine-based covalent organic frameworks as reticular ratchets to delicately manipulate the intrareticular charge transfer for directing ECL transduction from electric and chemical energies. Aligning the donor and acceptor (D-A) directions with the imine dipole effectively facilitates charge migration, whereas reversing the D-A direction impedes it. Notably, the ratchet effect of charge transfer directionality intensified with increasing D-A contrast, resulting in a remarkable 680-fold improvement in the ECL efficiency. Furthermore, dipole-controlled exciton binding energy, electron/hole decay kinetics, and femtosecond transient absorption spectra identified the electron transfer tendency from the N-end toward the C-end of reticular ratchets during ECL transduction. An exponential correlation between the ECL efficiency and the dipole difference was discovered. Our work provides a general approach to manipulate charge transfer and design next-generation electrochemical devices.

Green-Light-Triggered and Self-Calibrated Cascade Release of Nitric Oxide and Carbon Monoxide for Synergistic Glaucoma Therapy.

Glaucoma is an optic degenerative neuropathy that is driven by a vicious cycle of oxidative stress and mechanical stress. Current clinical treatments aim exclusively at alleviating mechanical stress by reducing the intraocular pressure (IOP). With the unattended oxidative stress, recurrence and deterioration of mechanical stress are inevitable. Nitric oxide (NO) and carbon monoxide (CO) are endogenous gaseous signaling molecules for vasodilation and anti-inflammation, respectively. Mounting evidence suggests an intricate interplay between NO and CO to mediate their biological roles, like how it takes two to dance a waltz. This leads to the concept of "gas waltz therapy" for glaucoma, in which NO is released to reduce IOP and stoichiometric CO is coreleased to suppress oxidative stress. CND570 is the first phototriggered cascade NO/CO donor, to the best of our knowledge. Notably, the release of NO/CO is accompanied by the concomitant release of a rhodamine dye whose bright fluorescence is harnessed as a convenient calibration mechanism of the gas release profile. CND570 exhibits excellent transcorneal permeability and reaches the target aqueous humor outflow pathway. Further, green-light irradiation triggers release of CO and NO in the eye tissue of glaucoma mice. NO and CO could promote the upregulation of soluble guanylate cyclase (sGC) in both in vitro and in vivo models. Notably, CND570 treatment significantly reduces the oxidative stress associated with glaucoma. NO/CO-based gas waltz therapy is a promising new avenue for glaucoma treatment.

Breaking Low-Strain and Deep-Potassiation Trade-Off in Alloy Anodes via Bonding Modulation for High-Performance K-Ion Batteries.

Alloy anode materials have garnered unprecedented attention for potassium storage due to their high theoretical capacity. However, the substantial structural strain associated with deep potassiation results in serious electrode fragmentation and inadequate K-alloying reactions. Effectively reconciling the trade-off between low-strain and deep-potassiation in alloy anodes poses a considerable challenge due to the larger size of K-ions compared to Li/Na-ions. In this study, we propose a chemical bonding modulation strategy through single-atom modification to address the volume expansion of alloy anodes during potassiation. Using black phosphorus (BP) as a representative and generalizing to other alloy anodes, we established a robust P-S covalent bonding network via sulfur doping. This network exhibits sustained stability across discharge-charge cycles, elevating the modulus of K-P compounds by 74%, effectively withstanding the high strain induced by the potassiation process. Additionally, the bonding modulation reduces the formation energies of potassium phosphides, facilitating a deeper potassiation of the BP anode. As a result, the modified BP anode exhibits a high reversible capacity and extended operational lifespan, coupled with a high areal capacity. This work introduces a new perspective on overcoming the trade-off between low-strain and deep-potassiation in alloy anodes for the development of high-energy and stable potassium-ion batteries.

Brain structure, amyloid, and behavioral features for predicting clinical progression in subjective cognitive decline.

As a potential preclinical stage of Alzheimer's dementia, subjective cognitive decline (SCD) reveals a higher risk of future cognitive decline and conversion to dementia. However, it has not been clear whether SCD status increases the clinical progression of older adults in the context of amyloid deposition, cerebrovascular disease (CeVD), and psychiatric symptoms. We identified 99 normal controls (NC), 15 SCD individuals who developed mild cognitive impairment in the next 2 years (P-SCD), and 54 SCD individuals who did not (S-SCD) from ADNI database with both baseline and 2-year follow-up data. Total white matter hyperintensity (WMH), WMH in deep (DWMH) and periventricular (PWMH) regions, and voxel-wise grey matter volumes were compared among groups. Furthermore, using structural equation modelling method, we constructed path models to explore SCD-related brain changes longitudinally and to determine whether baseline SCD status, age, and depressive symptoms affect participants' clinical outcomes. Both SCD groups showed higher baseline amyloid PET SUVR, baseline PWMH volumes, and larger increase of PWMH volumes over time than NC. In contrast, only P-SCD had higher baseline DWMH volumes and larger increase of DWMH volumes over time than NC. No longitudinal differences in grey matter volume and amyloid was observed among NC, S-SCD, and P-SCD. Our path models demonstrated that SCD status contributed to future WMH progression. Further, baseline SCD status increases the risk of future cognitive decline, mediated by PWMH; baseline depressive symptoms directly contribute to clinical outcomes. In conclusion, both S-SCD and P-SCD exhibited more severe CeVD than NC. The CeVD burden increase was more pronounced in P-SCD. In contrast with the direct association of depressive symptoms with dementia severity progression, the effects of SCD status on future cognitive decline may manifest via CeVD pathologies. Our work highlights the importance of multi-modal longitudinal designs in understanding the SCD trajectory heterogeneity, paving the way for stratification and early intervention in the preclinical stage. PRACTITIONER POINTS: Both S-SCD and P-SCD exhibited more severe CeVD at baseline and a larger increase of CeVD burden compared to NC, while the burden was more pronounced in P-SCD. Baseline SCD status increases the risk of future PWMH and DWMH volume accumulation, mediated by baseline PWMH and DWMH volumes, respectively. Baseline SCD status increases the risk of future cognitive decline, mediated by baseline PWMH, while baseline depression status directly contributes to clinical outcome.

Small vessel disease and cognitive reserve oppositely modulate global network redundancy and cognitive function: A study in middle-to-old aged community participants.

Cerebral small vessel disease (SVD) can disrupt the global brain network and lead to cognitive impairment. Conversely, cognitive reserve (CR) can improve one's cognitive ability to handle damaging effects like SVD, partly by optimizing the brain network's organization. Understanding how SVD and CR collectively influence brain networks could be instrumental in preventing cognitive impairment. Recently, brain redundancy has emerged as a critical network protective metric, providing a nuanced perspective of changes in network organization. However, it remains unclear how SVD and CR affect global redundancy and subsequently cognitive function. Here, we included 121 community-dwelling participants who underwent neuropsychological assessments and a multimodal MRI examination. We visually examined common SVD imaging markers and assessed lifespan CR using the Cognitive Reserve Index Questionnaire. We quantified the global redundancy index (RI) based on the dynamic functional connectome. We then conducted multiple linear regressions to explore the specific cognitive domains related to RI and the associations of RI with SVD and CR. We also conducted mediation analyses to explore whether RI mediated the relationships between SVD, CR, and cognition. We found negative correlations of RI with the presence of microbleeds (MBs) and the SVD total score, and a positive correlation of RI with leisure activity-related CR (CRI-leisure). RI was positively correlated with memory and fully mediated the relationships between the MBs, CRI-leisure, and memory. Our study highlights the potential benefits of promoting leisure activities and keeping brain redundancy for memory preservation in older adults, especially those with SVD.

HMGB1-RAGE axis contributes to myocardial ischemia/reperfusion injury via regulation of cardiomyocyte autophagy and apoptosis in diabetic mice.

Patients with acute myocardial infarction complicated with diabetes are more likely to develop myocardial ischemia/reperfusion (I/R) injury (MI/RI) during reperfusion therapy. Both HMGB1 and RAGE play important roles in MI/RI. However, the specific mechanisms of HMGB1 associated with RAGE are not fully clarified in diabetic MI/RI. This study aimed to investigate whether the HMGB1-RAGE axis induces diabetic MI/RI via regulating autophagy and apoptosis. A db/db mouse model of MI/RI was established, where anti-HMGB1 antibody and RAGE inhibitor (FPS-ZM1) were respectively injected after 10 min of reperfusion. The results showed that treatment with anti-HMGB1 significantly reduced the infarct size, serum LDH, and CK-MB level. Similar situations also occurred in mice administrated with FPS-ZM1, though the HMGB1 level was unchanged. Then, we found that treatment with anti-HMGB1 or FPS-ZM1 performed the same effects in suppressing the autophagy and apoptosis, as reflected by the results of lower LAMP2 and LC3B levels, increased Bcl-2 level, reduced BAX and caspase-3 levels. Moreover, the Pink1/Parkin levels were also inhibited at the same time. Collectively, this study indicates that the HMGB1-RAGE axis aggravated diabetic MI/RI via apoptosis and Pink1/Parkin mediated autophagy pathways, and inhibition of HMGB1 or RAGE contributes to alleviating those adverse situations.

DLRAPom: a hybrid pipeline of Optimized XGBoost-guided integrative multiomics analysis for identifying targetable disease-related lncRNA-miRNA-mRNA regulatory axes.

The lack of a reliable and easy-to-operate screening pipeline for disease-related noncoding RNA regulatory axis is a problem that needs to be solved urgently. To address this, we designed a hybrid pipeline, disease-related lncRNA-miRNA-mRNA regulatory axis prediction from multiomics (DLRAPom), to identify risk biomarkers and disease-related lncRNA-miRNA-mRNA regulatory axes by adding a novel machine learning model on the basis of conventional analysis and combining experimental validation. The pipeline consists of four parts, including selecting hub biomarkers by conventional bioinformatics analysis, discovering the most essential protein-coding biomarkers by a novel machine learning model, extracting the key lncRNA-miRNA-mRNA axis and validating experimentally. Our study is the first one to propose a new pipeline predicting the interactions between lncRNA and miRNA and mRNA by combining WGCNA and XGBoost. Compared with the methods reported previously, we developed an Optimized XGBoost model to reduce the degree of overfitting in multiomics data, thereby improving the generalization ability of the overall model for the integrated analysis of multiomics data. With applications to gestational diabetes mellitus (GDM), we predicted nine risk protein-coding biomarkers and some potential lncRNA-miRNA-mRNA regulatory axes, which all correlated with GDM. In those regulatory axes, the MALAT1/hsa-miR-144-3p/IRS1 axis was predicted to be the key axis and was identified as being associated with GDM for the first time. In short, as a flexible pipeline, DLRAPom can contribute to molecular pathogenesis research of diseases, effectively predicting potential disease-related noncoding RNA regulatory networks and providing promising candidates for functional research on disease pathogenesis.

Improving protein structure prediction using templates and sequence embedding.

MOTIVATION: Protein structure prediction has been greatly improved by deep learning, but the contribution of different information is yet to be fully understood. This article studies the impacts of two kinds of information for structure prediction: template and multiple sequence alignment (MSA) embedding. Templates have been used by some methods before, such as AlphaFold2, RoseTTAFold and RaptorX. AlphaFold2 and RosetTTAFold only used templates detected by HHsearch, which may not perform very well on some targets. In addition, sequence embedding generated by pre-trained protein language models has not been fully explored for structure prediction. In this article, we study the impact of templates (including the number of templates, the template quality and how the templates are generated) on protein structure prediction accuracy, especially when the templates are detected by methods other than HHsearch. We also study the impact of sequence embedding (generated by MSATransformer and ESM-1b) on structure prediction. RESULTS: We have implemented a deep learning method for protein structure prediction that may take templates and MSA embedding as extra inputs. We study the contribution of templates and MSA embedding to structure prediction accuracy. Our experimental results show that templates can improve structure prediction on 71 of 110 CASP13 (13th Critical Assessment of Structure Prediction) targets and 47 of 91 CASP14 targets, and templates are particularly useful for targets with similar templates. MSA embedding can improve structure prediction on 63 of 91 CASP14 (14th Critical Assessment of Structure Prediction) targets and 87 of 183 CAMEO targets and is particularly useful for proteins with shallow MSAs. When both templates and MSA embedding are used, our method can predict correct folds (TMscore > 0.5) for 16 of 23 CASP14 FM targets and 14 of 18 Continuous Automated Model Evaluation (CAMEO) targets, outperforming RoseTTAFold by 5% and 7%, respectively. AVAILABILITY AND IMPLEMENTATION: Available at https://github.com/xluo233/RaptorXFold. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

Hierarchical graph transformer with contrastive learning for protein function prediction.

MOTIVATION: In recent years, high-throughput sequencing technologies have made large-scale protein sequences accessible. However, their functional annotations usually rely on low-throughput and pricey experimental studies. Computational prediction models offer a promising alternative to accelerate this process. Graph neural networks have shown significant progress in protein research, but capturing long-distance structural correlations and identifying key residues in protein graphs remains challenging. RESULTS: In the present study, we propose a novel deep learning model named Hierarchical graph transformEr with contrAstive Learning (HEAL) for protein function prediction. The core feature of HEAL is its ability to capture structural semantics using a hierarchical graph Transformer, which introduces a range of super-nodes mimicking functional motifs to interact with nodes in the protein graph. These semantic-aware super-node embeddings are then aggregated with varying emphasis to produce a graph representation. To optimize the network, we utilized graph contrastive learning as a regularization technique to maximize the similarity between different views of the graph representation. Evaluation of the PDBch test set shows that HEAL-PDB, trained on fewer data, achieves comparable performance to the recent state-of-the-art methods, such as DeepFRI. Moreover, HEAL, with the added benefit of unresolved protein structures predicted by AlphaFold2, outperforms DeepFRI by a significant margin on Fmax, AUPR, and Smin metrics on PDBch test set. Additionally, when there are no experimentally resolved structures available for the proteins of interest, HEAL can still achieve better performance on AFch test set than DeepFRI and DeepGOPlus by taking advantage of AlphaFold2 predicted structures. Finally, HEAL is capable of finding functional sites through class activation mapping. AVAILABILITY AND IMPLEMENTATION: Implementations of our HEAL can be found at https://github.com/ZhonghuiGu/HEAL.