ProtVar: mapping and contextualizing human missense variation.

Genomic variation can impact normal biological function in complex ways and so understanding variant effects requires a broad range of data to be coherently assimilated. Whilst the volume of human variant data and relevant annotations has increased, the corresponding increase in the breadth of participating fields, standards and versioning mean that moving between genomic, coding, protein and structure positions is increasingly complex. In turn this makes investigating variants in diverse formats and assimilating annotations from different resources challenging. ProtVar addresses these issues to facilitate the contextualization and interpretation of human missense variation with unparalleled flexibility and ease of accessibility for use by the broadest range of researchers. By precalculating all possible variants in the human proteome it offers near instantaneous mapping between all relevant data types. It also combines data and analyses from a plethora of resources to bring together genomic, protein sequence and function annotations as well as structural insights and predictions to better understand the likely effect of missense variation in humans. It is offered as an intuitive web server https://www.ebi.ac.uk/protvar where data can be explored and downloaded, and can be accessed programmatically via an API.

Reduced mitochondrial pyruvate carrier expression in hearts with heart failure and reduced ejection fraction patients: ischemic vs. non-ischemic origin.

Introduction and objectives: Mitochondrial pyruvate carrier (MPC) mediates the entry of pyruvate into mitochondria, determining whether pyruvate is incorporated into the Krebs cycle or metabolized in the cytosol. In heart failure (HF), a large amount of pyruvate is metabolized to lactate in the cytosol rather than being oxidized inside the mitochondria. Thus, MPC activity or expression might play a key role in the fate of pyruvate during HF. The purpose of this work was to study the levels of the two subunits of this carrier, named MPC1 and MPC2, in human hearts with HF of different etiologies. Methods: Protein and mRNA expression analyses were conducted in cardiac tissues from three donor groups: patients with HF with reduced ejection fraction (HFrEF) with ischemic cardiomyopathy (ICM) or idiopathic dilated cardiomyopathy (IDC), and donors without cardiac pathology (Control). MPC2 plasma levels were determined by ELISA. Results: Significant reductions in the levels of MPC1, MPC2, and Sirtuin 3 (SIRT3) were observed in ICM patients compared with the levels in the Control group. However, no statistically significant differences were revealed in the analysis of MPC1 and MPC2 gene expression among the groups. Interestingly, Pyruvate dehydrogenase complex (PDH) subunits expression were increased in the ICM patients. In the case of IDC patients, a significant decrease in MPC1 was observed only when compared with the Control group. Notably, plasma MPC2 levels were found to be elevated in both disease groups compared with that in the Control group. Conclusion: Decreases in MPC1 and/or MPC2 levels were detected in the cardiac tissues of HFrEF patients, with ischemic or idiopatic origen, indicating a potential reduction in mitochondrial pyruvate uptake in the heart, which could be linked to unfavorable clinical features.

Inflammatory bowel disease induces pathological α-synuclein aggregation in the human gut and brain.

AIMS: According to Braak's hypothesis, it is plausible that Parkinson's disease (PD) originates in the enteric nervous system (ENS) and spreads to the brain through the vagus nerve. In this work, we studied whether inflammatory bowel diseases (IBDs) in humans can progress with the emergence of pathogenic α-synuclein (α-syn) in the gastrointestinal tract and midbrain dopaminergic neurons. METHODS: We have analysed the gut and the ventral midbrain from subjects previously diagnosed with IBD and form a DSS-based rat model of gut inflammation in terms of α-syn pathology. RESULTS: Our data support the existence of pathogenic α-syn in both the gut and the brain, thus reinforcing the potential role of the ENS as a contributing factor in PD aetiology. Additionally, we have analysed the effect of a DSS-based rat model of gut inflammation to demonstrate (i) the appearance of P-α-syn inclusions in both Auerbach's and Meissner's plexuses (gut), (ii) an increase in α-syn expression in the ventral mesencephalon (brain) and (iii) the degeneration of nigral dopaminergic neurons, which all are considered classical hallmarks in PD. CONCLUSION: These results strongly support the plausibility of Braak's hypothesis and emphasise the significance of peripheral inflammation and the gut-brain axis in initiating α-syn aggregation and transport to the substantia nigra, resulting in neurodegeneration.

Development and validation of a software application to analyze thermal and kinematic multimodels of Stirling engines.

The work developed presents, for the first time, a tool to analyze all the thermodynamic models used in the study and development of Stirling engines: isothermal, ideal adiabatic and adiabatic with losses, combined adiabatic thermodynamic with finite speed (CAFS), thermodynamic with finite speed (FST), ideal polytropic and polytropic with losses (PSVL), allowing a comparative study of them. This software (ASCE-UMA), designed and implemented in a Matlab GUI® allows to obtain the operating parameters of these engines, calculating the thermodynamic parameters, power output and efficiency. Additionally, the thermodynamic models can be evaluated with different mechanical configurations, for which different drive mechanisms are implemented: Sinusoidal, Alfa Ross yoke types, Alfa Ross V yoke, Beta rhombic type and free piston Stirling engine (FPSE). Thermoacoustic and other, models could be analyzed by virtue of their similarity of movement with some of the implemented models. In the same way, ASCE-UMA allows the study of various exchanger configurations, as well as various regenerator models. The versatility of ASCE-UMA allows the development analysis of all the fundamental elements of a new prototype as well as the analysis of experimental data by performing a customized and detailed calculation. To test the effectiveness of ASCE-UMA, its performance is verified by analyzing Ross Yoke D-90 models and a GM GPU-3 engine. This is a tool that allows to analyze and comparing the different models and the different existing mechanisms for the multiple configurations of Stirling engines in an easy and intuitive application with a high-quality graphical interface.

The effects of pathogenic and likely pathogenic variants for inherited hemostasis disorders in 140 214 UK Biobank participants.

Rare genetic diseases affect millions, and identifying causal DNA variants is essential for patient care. Therefore, it is imperative to estimate the effect of each independent variant and improve their pathogenicity classification. Our study of 140 214 unrelated UK Biobank (UKB) participants found that each of them carries a median of 7 variants previously reported as pathogenic or likely pathogenic. We focused on 967 diagnostic-grade gene (DGG) variants for rare bleeding, thrombotic, and platelet disorders (BTPDs) observed in 12 367 UKB participants. By association analysis, for a subset of these variants, we estimated effect sizes for platelet count and volume, and odds ratios for bleeding and thrombosis. Variants causal of some autosomal recessive platelet disorders revealed phenotypic consequences in carriers. Loss-of-function variants in MPL, which cause chronic amegakaryocytic thrombocytopenia if biallelic, were unexpectedly associated with increased platelet counts in carriers. We also demonstrated that common variants identified by genome-wide association studies (GWAS) for platelet count or thrombosis risk may influence the penetrance of rare variants in BTPD DGGs on their associated hemostasis disorders. Network-propagation analysis applied to an interactome of 18 410 nodes and 571 917 edges showed that GWAS variants with large effect sizes are enriched in DGGs and their first-order interactors. Finally, we illustrate the modifying effect of polygenic scores for platelet count and thrombosis risk on disease severity in participants carrying rare variants in TUBB1 or PROC and PROS1, respectively. Our findings demonstrate the power of association analyses using large population datasets in improving pathogenicity classifications of rare variants.

Mulching vs. organic soil amendment: Effects on adsorption-desorption of herbicides.

Mulching and organic soil amendment are two agricultural practices that are being increasingly used to preserve soil from degradation, although they may modify the fate of herbicides when applied in soils subjected to these practices. This study has set out to compare the impact of both agricultural practices on the adsorption-desorption behaviour of the herbicides S-metolachlor (SMOC), foramsulfuron (FORAM), and thiencarbazone-methyl (TCM) involving winter wheat mulch residues at different stages of decomposition and particle size, and unamended soils or those amended with mulch. The Freundlich Kf adsorption constants of the three herbicides by mulches, and unamended and amended soils ranged between 1.34 and 65.8 (SMOC), 0-34.3 (FORAM), and 0.01-1.10 (TCM). The adsorption of the three compounds was significantly higher in mulches than in soils (unamended and amended). The adsorption of SMOC and FORAM increased significantly with mulch decomposition, with this positive impact also being observed on the adsorption of FORAM and TCM after mulch milling. Simple and multiple correlations between mulches, soils, and herbicide properties, and adsorption-desorption constants (Kf, Kd, Kfd) reflected the organic carbon (OC) content and/or dissolved organic carbon (DOC) content of the adsorbents as main variables controlling the adsorption and/or desorption of each herbicide. The statistic R2 revealed that >61 % of the variability in the adsorption-desorption constants could be explained by jointly considering the OC of mulches and soils and the hydrophobicity (for Kf) or water solubility of herbicides (for Kd or Kfd). The same trend observed for Kfd desorption constants as for Kf adsorption ones resulted in higher percentages of herbicide remaining adsorbed after desorption in amended soils (33 %-41 % of SMOC, 0 %-15 % of FORAM, and 2 %-17 % of TCM) than in mulches (< 10 %). The results reveal a higher efficiency of organic soil amendment than mulching as an agricultural practice for immobilising the herbicides studied when winter wheat mulch residues are used as a common adsorbent, and as a better strategy for avoiding groundwater contamination.

Transfer learning for drug-target interaction prediction.

MOTIVATION: Utilizing AI-driven approaches for drug-target interaction (DTI) prediction require large volumes of training data which are not available for the majority of target proteins. In this study, we investigate the use of deep transfer learning for the prediction of interactions between drug candidate compounds and understudied target proteins with scarce training data. The idea here is to first train a deep neural network classifier with a generalized source training dataset of large size and then to reuse this pre-trained neural network as an initial configuration for re-training/fine-tuning purposes with a small-sized specialized target training dataset. To explore this idea, we selected six protein families that have critical importance in biomedicine: kinases, G-protein-coupled receptors (GPCRs), ion channels, nuclear receptors, proteases, and transporters. In two independent experiments, the protein families of transporters and nuclear receptors were individually set as the target datasets, while the remaining five families were used as the source datasets. Several size-based target family training datasets were formed in a controlled manner to assess the benefit provided by the transfer learning approach. RESULTS: Here, we present a systematic evaluation of our approach by pre-training a feed-forward neural network with source training datasets and applying different modes of transfer learning from the pre-trained source network to a target dataset. The performance of deep transfer learning is evaluated and compared with that of training the same deep neural network from scratch. We found that when the training dataset contains fewer than 100 compounds, transfer learning outperforms the conventional strategy of training the system from scratch, suggesting that transfer learning is advantageous for predicting binders to under-studied targets. AVAILABILITY AND IMPLEMENTATION: The source code and datasets are available at https://github.com/cansyl/TransferLearning4DTI. Our web-based service containing the ready-to-use pre-trained models is accessible at https://tl4dti.kansil.org.

Searching and Navigating UniProt Databases.

The Universal Protein Resource (UniProt) is a comprehensive resource for protein sequence and annotation data. The UniProt website receives about 800,000 unique visitors per month and is the primary means to access UniProt. It provides 10 searchable datasets and four main tools. The key UniProt datasets are the UniProt Knowledgebase (UniProtKB), the UniProt Reference Clusters (UniRef), the UniProt Archive (UniParc), and protein sets for completely sequenced genomes (Proteomes). Other supporting datasets include information about proteins that is present in UniProtKB protein entries, such as literature citations, taxonomy, and subcellular locations, among others. This article focuses on how to use UniProt datasets. The first basic protocol describes navigation and searching mechanisms for the UniProt datasets, and two additional protocols build on the first protocol to describe advanced search and query building. © 2023 The Authors. Current Protocols published by Wiley Periodicals LLC. Basic Protocol 1: Searching UniProt datasets Basic Protocol 2: Advanced search and query building Basis Protocol 3: Adding parameters using advanced search.

RNY3 modulates cell proliferation and IL13 mRNA levels in a T lymphocyte model: a possible new epigenetic mechanism of IL-13 regulation

Allergic asthma is the most common type of asthma. It is characterized by TH2 cell–driven inflammation in which interleukin-13 (IL-13) plays a pivotal role. Cytoplasmic RNAs (Y-RNAs), a variety of non-coding RNAs that are dysregulated in many cancer types, are also differentially expressed in patients with allergic asthma. Their function in the development of the disease is still unknown. We investigated the potential role of RNY3 RNA (hY3) in the TH2 cell inflammatory response using the Jurkat cell line as a model. hY3 expression levels were modulated to mimic the upregulation effect in allergic disease. We evaluated the effect of hY3 over cell stimulation and the expression of the TH2 cytokine IL13. Total RNA was isolated and retrotranscribed, and RNA levels were assessed by qPCR. In Jurkat cells, hY3 levels increased upon stimulation with phorbol 12-myristate 13-acetate (PMA) and ionomycin. When transfecting with high levels of hY3 mimic molecules, cell proliferation rate decreased while IL13 mRNA levels increased upon stimulation compared to stimulated control cells. Our results show the effect of increased hY3 levels on cell proliferation and the levels of IL13 mRNA in Jurkat cells. Also, we showed that hY3 could act over other cells via exosomes. This study opens up new ways to study the potential regulatory function of hY3 over IL-13 production and its implications for asthma development. (AU)

Changes in vineyard soil parameters after repeated application of organic-inorganic amendments based on spent mushroom substrate.

The changes of physicochemical and biochemical parameters of a silty loam (S1) and sandy loam (S2) vineyard soils added with spent mushroom substrate (SMS) or SMS composted with ophite (OF) as rock dust (SMS + OF) were studied. Two doses of SMS or SMS + OF (25 and 100 Mg ha-1) were applied for two consecutive years (2020-2021) and changes of soil physicochemical parameters, and dehydrogenase activity (DHA), respiration (RES), microbial biomass (BIO), and the phospholipid fatty acids (PLFAs) profile were assayed on a temporal basis. The results showed an increase in soil organic carbon (OC) content, total and mineralised N, P, and K, especially when the highest SMS dose was applied to soils. Repeated application caused OC content over time up to 2.3 times higher than initial content in the silty loam soil. This increase was not observed in sandy soil, possibly due to a higher bioavailability of OC, as indicated by the evolution of extractable humic acid/fulvic acid pools. In both soils, all biochemical parameters increased after amendment, being favoured both by the OC and by the presence of OF. Significant positive correlations were found between DHA, RES and BIO, and OC content especially in the first part and then levelled off after the second dose application. Total bacterial or fungal PLFAs patterns reflected the variation of BIO by SMS application. The higher growth of fungi vs. bacterial community in amended soils was recorded after the first SMS application, although the opposite effect occurred after the second application, with similar results in both soils. The findings indicate that the application of SMS or SMS + OF in vineyard soils could be an appropriate agronomic management practice for maintaining soil sustainability, although doses and application times of these amendments should first be evaluated depending on soil texture.

Soluble HLA-G levels in heart transplant recipients: Dynamics and correlation with clinical outcomes.

PURPOSE: To describe the evolution of the serum levels of soluble HLA-G (s-HLA-G) during the first 12 months after heart transplantation (HT) and to correlate it with clinical outcomes. METHODS: Observational study based in a single-center cohort of 59 patients who underwent HT between December-2003 and March-2010. Soluble HLA-G levels were measured from serum samples extracted before HT, and 1, 3, 6 and 12 months after HT. The cumulative burden of s-HLA-G expression during the first post-transplant year was assessed by means of the area under the curve (AUC) of s-HLA-G levels over time and correlated with the acute rejection burden -as assessed by a rejection score-, the presence of coronary allograft vasculopathy (CAV) grade ≥ 1 and infections during the first post-transplant year; as well as with long-term patient and graft survival. Mean follow-up was 12.4 years. RESULTS: Soluble HLA-G levels decreased over the first post-transplant year (p = 0.020). The AUC of s-HLA-G levels during the first post-transplant year was higher among patients with infections vs. those without infections (p = 0.006). No association was found between the AUC of s-HLA-G levels and the burden of acute rejection or the development of CAV. Overall long-term survival, long-term survival free of late graft failure and cancer-free survival were not significantly different in patients with an AUC of s-HLA-G levels higher or lower than the median of the study population. CONCLUSIONS: Soluble HLA-G levels decreased over the first year after HT. Higher HLA-G expression was associated with a higher frequency of infections, but not with the burden of acute rejection or the development of CAV, neither with long-term patient or graft survival.

Laboratory interpretative comments and guidance: clinical and operative outcomes on moderate to severe hyponatraemia patient management.

AIMS: Hyponatraemia is the most common body fluid disorders but often goes unnoticed. Our laboratory incorporated a standardised procedure to help clinicians detect moderate/severe hyponatraemia. The study aims were to evaluate the outcomes on patient care and clinicians' satisfaction. METHODS: The study, observational and retrospective, included 1839 cases, adult and paediatric patients, with sodium concentration <130 mmol/L. The procedure consisted of interpretative comments in the emergency and core laboratories report and the point-of-care testing blood gas network report. We evaluated hyponatraemia length in two equal periods: before and after the implementation. We conducted a survey addressed to the staff of the clinical settings involved to know their satisfaction. RESULTS: The median hyponatraemia length decreased significantly from 4.95 hours (2.08-16.57) in the first period to 2.17 hours (1.06-5.39) in the second period. The lack of hyponatraemia patients follow-up was significantly less after the procedure implementation. The survey was answered by 92 (60 senior specialists and 32 residents) out of 110 clinicians surveyed. Ninety of them (98%) answered positively. CONCLUSIONS: We have demonstrated the reduction in the time for diagnosing and management by physicians, the higher uniformity in the time required to solve hyponatraemia episodes following our laboratory procedure and the clinicians' satisfaction.

Combining participatory action research and emerging ways of collective action to promote institutional change toward social commitment: Groundings, strategies, and implications of an experience.

AIMS: This study reports the foundations, strategies, and results of an institutional change experience based on the combination of participatory-action-research and new currents of collective mobilization and political participation. It aimed to achieve the institution's greater social commitment and a more participatory and transparent management. METHODS: The process took place in a Spanish public university and was promoted and coordinated by a Work Group that emerged from grassroots university community. Collective diagnosis was performed through face-to-face strategies (global, sectorial, and faculty meetings) and virtual tools (web-blog, on-line surveys, shared documents). Collective action combined nonformal with formal institutional participation and applied hybrid activism, self-organization in horizontal structures and integrative conflict management. RESULTS: A sequential process of diagnosis, collective action, and negotiation was implemented. As a result, the university Governing Team, representatives from different sectors and members of the Work Group worked jointly to define several institutional actions that were thereafter launched. Those actions aimed to improve institutional participation and transparency, and greater institutional social commitment. CONCLUSION: The combination of participatory-action-research and new ways of collective action can be an excellent tool to draw institutions towards greater social engagement, thus contributing to sustainable social change. A model to guide institutional change is drafted.

RNY3 modulates cell proliferation and IL13 mRNA levels in a T lymphocyte model: a possible new epigenetic mechanism of IL-13 regulation.

Allergic asthma is the most common type of asthma. It is characterized by TH2 cell-driven inflammation in which interleukin-13 (IL-13) plays a pivotal role. Cytoplasmic RNAs (Y-RNAs), a variety of non-coding RNAs that are dysregulated in many cancer types, are also differentially expressed in patients with allergic asthma. Their function in the development of the disease is still unknown. We investigated the potential role of RNY3 RNA (hY3) in the TH2 cell inflammatory response using the Jurkat cell line as a model. hY3 expression levels were modulated to mimic the upregulation effect in allergic disease. We evaluated the effect of hY3 over cell stimulation and the expression of the TH2 cytokine IL13. Total RNA was isolated and retrotranscribed, and RNA levels were assessed by qPCR. In Jurkat cells, hY3 levels increased upon stimulation with phorbol 12-myristate 13-acetate (PMA) and ionomycin. When transfecting with high levels of hY3 mimic molecules, cell proliferation rate decreased while IL13 mRNA levels increased upon stimulation compared to stimulated control cells. Our results show the effect of increased hY3 levels on cell proliferation and the levels of IL13 mRNA in Jurkat cells. Also, we showed that hY3 could act over other cells via exosomes. This study opens up new ways to study the potential regulatory function of hY3 over IL-13 production and its implications for asthma development.

Prevalence of Tricuspid Regurgitation After Orthotopic Heart Transplantation and Its Evolution in the Follow-up Period: A Long-Term Study.

BACKGROUND: Tricuspid regurgitation (TR) after heart transplant (HT) can be an important complication depending on its etiology and severity. This study aims to analyze the prevalence of TR, the causes, and its evolution over time after HT. METHODS: We performed a retrospective study of transplants performed between 2000 and 2019 in 2 centers (1009 patients). TR was grouped according to etiology: primary graft dysfunction (PGD), acute rejection, cardiac allograft vasculopathy (CAV), pulmonary hypertension, prolapse, endomyocardial biopsy complication (EMB), pacemaker (PM), and unclear etiology (TR not related to any process and for which no justification was found). RESULTS: The prevalence of TR after HT was 19.8% (moderate: 13.2%, severe: 6.6%). Significant TR was more prevalent in the first months (month 1: 51%, month 3: 40%, month 6: 29%, 1 year: 24%). These results were related to the etiologies. Thus, in the first month, TR due to PGD is frequent and it is the only time when TR due to pulmonary hypertension appears. During the first 6 months, TR of unclear cause gains relevance, which tends to decrease over time. After 1 year, TR due to rejection predominates. After 5 years, TR is less frequent (< 10%) and related to long-term complications of HT, such as CAV, EMB, and those associated with PM. CONCLUSIONS: The prevalence of TR after HT is 19.8%. Prevalence and etiology change over time. Initially it is usually related to PGD, in the medium-term to rejection and in the long-term to CAV and procedures such as EMB and PM.

NOM fractionation by HPSEC-DAD-OCD for predicting trihalomethane disinfection by-product formation potential in full-scale drinking water treatment plants.

Chlorination is a common method for water disinfection; however, it leads to the formation of disinfection by-products (DBPs), which are undesirable toxic pollutants. To prevent their formation, it is crucial to understand the reactivity of natural organic matter (NOM), which is considered a dominant precursor of DBPs. We propose a novel size exclusion chromatography (SEC) approach to evaluate NOM reactivity and the formation potential of total trihalomethanes-formation potentials (tTHMs-FP) and four regulated species (i.e. CHCl3, CHBrCl2, CHBr2Cl, and CHBr3). This method combines enhanced SEC separation with two analytical columns working in tandem and quantification of apparent molecular weight (AMW) NOM fractions using C content (organic carbon detector, OCD), 254-nm spectroscopic (diode-array detector, DAD) measurements, and spectral slopes at low (S206-240) and high (S350-380) wavelengths. Links between THMs-FP and NOM fractions from high performance size exclusion chromatography HPSEC-DAD-OCD were investigated using statistical modelling with multiple linear regressions for samples taken alongside conventional full-scale as well as full- and pilot-scale electrodialysis reversal and bench-scale ion exchange resins. The proposed models revealed promising correlations between the AMW NOM fractions and the THMs-FP. Methodological changes increased fractionated signal correlations relative to bulk regressions, especially in the proposed HPSEC-DAD-OCD method. Furthermore, spectroscopic models based on fractionated signals are presented, providing a promising approach to predict THMs-FP simultaneously considering the effect of the dominant THMs precursors, NOM and Br-.

Cancer in patients with heart failure: Incidence, risk factors and prognostic impact.

AIMS: To assess the incidence of cancer diagnosis and cancer-related mortality in patients with heart failure (HF). METHODS: Observational study based in a prospective cohort of patients with HF referred to a specialized Spanish clinic between 2010 and 2019. The observed incidence of malignancies (excluding non-melanoma skin cancer) was compared to that expected for the general Spanish population according to the Global Cancer Observatory. RESULTS: We studied 1909 consecutive patients with HF. Over a median follow-up of 4.07 years, 165 new cases of malignancy were diagnosed. Observed age-standardized incidence rates of cancer were 861 (95% CI 618.4-2159.4) cases per 100,000 patients-years in men and 728.5 (95% CI 451.1-4308.7) cases per 100,000 patients-years in women; while age-standardized incidence rates of cancer expected for the general Spanish population were 479.4 cases per 100,000 patients-years in men (risk ratio = 1.80) and 295.5 cases per 100,000 patients-years in women (risk ratio = 2.46). Both a history of pre-existing malignancy at baseline and the development of new malignancies during follow-up were associated with reduced survival. Observed age-standardized cancer-related mortality was 344.1 (95% CI 202.1-1675) deaths per 100,000 patient-years in men and 217.0 (95% CI 32.8-3949.3) deaths per 100,000 patient-years in women; while age-standardized cancer-related mortality expected for the general Spanish population was 201.4 deaths per 100,000 patients-years in men (risk ratio = 1.71) and 96.2 deaths per 100,000 patients-years in women (risk ratio = 2.26). CONCLUSION: Patients with HF showed higher incidence rates of cancer diagnosis and cancer-related mortality than those expected for the general population.

Hierarchical deep learning for predicting GO annotations by integrating protein knowledge.

MOTIVATION: Experimental testing and manual curation are the most precise ways for assigning Gene Ontology (GO) terms describing protein functions. However, they are expensive, time-consuming and cannot cope with the exponential growth of data generated by high-throughput sequencing methods. Hence, researchers need reliable computational systems to help fill the gap with automatic function prediction. The results of the last Critical Assessment of Function Annotation challenge revealed that GO-terms prediction remains a very challenging task. Recent developments on deep learning are significantly breaking out the frontiers leading to new knowledge in protein research thanks to the integration of data from multiple sources. However, deep models hitherto developed for functional prediction are mainly focused on sequence data and have not achieved breakthrough performances yet. RESULTS: We propose DeeProtGO, a novel deep-learning model for predicting GO annotations by integrating protein knowledge. DeeProtGO was trained for solving 18 different prediction problems, defined by the three GO sub-ontologies, the type of proteins, and the taxonomic kingdom. Our experiments reported higher prediction quality when more protein knowledge is integrated. We also benchmarked DeeProtGO against state-of-the-art methods on public datasets, and showed it can effectively improve the prediction of GO annotations. AVAILABILITY AND IMPLEMENTATION: DeeProtGO and a case of use are available at https://github.com/gamerino/DeeProtGO. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.