Mapping single-cell developmental potential in health and disease with interpretable deep learning.

Single-cell RNA sequencing (scRNA-seq) has transformed our understanding of cell fate in developmental systems. However, identifying the molecular hallmarks of potency - the capacity of a cell to differentiate into other cell types - has remained challenging. Here, we introduce CytoTRACE 2, an interpretable deep learning framework for characterizing potency and differentiation states on an absolute scale from scRNA-seq data. Across 31 human and mouse scRNA-seq datasets encompassing 28 tissue types, CytoTRACE 2 outperformed existing methods for recovering experimentally determined potency levels and differentiation states covering the entire range of cellular ontogeny. Moreover, it reconstructed the temporal hierarchy of mouse embryogenesis across 62 timepoints; identified pan-tissue expression programs that discriminate major potency levels; and facilitated discovery of cellular phenotypes in cancer linked to survival and immunotherapy resistance. Our results illuminate a fundamental feature of cell biology and provide a broadly applicable platform for delineating single-cell differentiation landscapes in health and disease.

Distinguishing SARS-CoV-2 infection and vaccine responses up to 18 months post-infection using nucleocapsid protein and receptor-binding domain antibodies.

The prediction of the durability of immunity against COVID-19 is relevant, and longitudinal studies are essential for unraveling the details regarding protective SARS-CoV-2 antibody responses. It has become challenging to discriminate between COVID-19 vaccine- and infection-induced immune responses since all approved vaccines in Europe and the USA are based on the viral spike (S) protein, which is also the most commonly used antigen in immunoassays measuring immunoglobulins (Igs) against SARS-CoV-2. We have developed a nucleocapsid (N) protein-based sandwich ELISA for detecting pan anti-SARS-CoV-2 Ig with a sensitivity and specificity of 97%. Generalized mixed models were used to determine the degree of long-term humoral immunity against the N protein and the receptor-binding domain (RBD) of the S protein in a cohort of infected individuals to distinguish between COVID-19 vaccine- and infection-induced immunity. N-specific waning could be observed in individuals who did not experience reinfection, while individuals who experienced reinfection had a new significant increase in N-specific Ig levels. In individuals that seroconverted without a reinfection, 70.1% remained anti-N seropositive after 550 days. The anti-RBD Ig dynamics were unaffected by reinfection but exhibited a clear increase in RBD-specific Ig when vaccination was initiated. In conclusion, a clear difference in the dynamics of the antibody response against N protein and RBD was observed over time. Anti-N protein-specific Igs can be detected up to 18 months after SARS-CoV-2 infection allowing long-term discrimination of infectious and vaccine antibody responses.IMPORTANCELongitudinal studies are essential to unravel details regarding the protective antibody responses after COVID-19 infection and vaccination. It has become challenging to distinguish long-term immune responses to SARS-CoV-2 infection and vaccination since most approved vaccines are based on the viral spike (S) protein, which is also mostly used in immunoassays measuring immunoglobulins (Igs) against SARS-CoV-2. We have developed a novel nucleocapsid (N) protein-based sandwich ELISA for detecting pan-anti-SARS-CoV-2 Ig, exhibiting high sensitivity and specificity. Generalized mixed models were used to determine long-term humoral immunity in a cohort of infected individuals from the Faroe Islands, distinguishing between COVID-19 vaccine- and infection-induced immunity. A clear difference in the dynamics of the antibody response against N protein and S protein was observed over time, and the anti-N protein-specific Igs could be detected up to 18 months after SARS-CoV-2 infection. This enables long-term discrimination between natural infection and vaccine-dependent antibody responses.

Humoral and cellular immune responses after three or four doses of BNT162b2 in patients with hematological malignancies.

OBJECTIVES: Initial responses to coronavirus disease 2019 vaccination are impaired in patients with hematological malignancies. We investigated immune responses after three or four doses of BNT162b2 in patients with myeloid and lymphoid malignancies compared to controls, and identified risk factors for humoral and cellular nonresponse 1 year after first vaccination. METHODS: In 407 hematological patients (45 myeloid, 362 lymphoid) and 98 matched controls, we measured immunoglobulin G (IgG) and neutralizing antibodies specific for the receptor-binding domain of severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) at baseline, 3 weeks, 2, 6, and 12 months, and interferon-γ release at 12 months. RESULTS: In patients with lymphoid malignancies, SARS-CoV-2 receptor-binding domain IgG concentration and mean neutralizing capacity was lower than in controls at all time points. A diagnosis of chronic lymphocytic B-cell leukemia (CLL) or lymphoma was associated with humoral nonresponse at 12 months compared to having multiple myeloma/amyloidosis (p < .001 and p = .013). Compared to controls, patients with lymphoid malignancies had increased risk of cellular nonresponse. A lymphoma diagnosis was associated with lower risk of cellular nonresponse compared to patients with multiple myeloma/amyloidosis, while patients with CLL had comparable response rates to patients with multiple myeloma/amyloidosis (p = .037 and p = .280). CONCLUSIONS: In conclusion, long-term humoral and cellular immune responses to BNT162b2 were impaired in patients with lymphoid malignancies.

Waning humoral and cellular immunity after COVID-19 vaccination in patients with psoriasis treated with methotrexate and biologics: a cohort study.

BACKGROUND: mRNA-based COVID-19 vaccines have short- and long-term efficacy in healthy individuals, but their efficacy in patients with psoriasis receiving immunomodulatory therapy is less studied. OBJECTIVES: To investigate long-term immunity after COVID-19 vaccination in patients with psoriasis receiving immunomodulatory therapy. METHODS: A prospective cohort study including patients (n = 123) with psoriasis receiving methotrexate (MTX) or biologics and controls (n = 226). Only mRNA-based COVID-19 vaccines administered with standard intervals between doses were investigated. Markers of immunity included SARS-CoV-2 spike glycoprotein-specific IgG and IgA, neutralizing capacity, and interferon-γ release from T cells stimulated with peptides of the SARS-CoV-2 spike glycoprotein. RESULTS: The proportion of IgG responders was lower 6 months after vaccination in patients receiving anti-tumour necrosis factor (TNF) treatment compared with controls. Anti-TNF treatment was associated with lower IgG levels (ß = -0.82, 95% confidence interval -1.38 to -0.25; P = 0.001). The median neutralizing index was lower in the anti-TNF group [50% inhibition (interquartile range [IQR] 37-89)] compared with controls [98% inhibition (IQR 96-99)]; P < 0.001. Cellular responses were numerically lowest in the anti-TNF group. CONCLUSIONS: Treatment with anti-TNF has an impact on the immunity elicited by mRNA-based COVID-19 vaccination in patients with psoriasis, resulting in a faster waning of humoral and cellular markers of immunity; however, the clinical implications are unknown.

Antibody responses and risk factors associated with impaired immunological outcomes following two doses of BNT162b2 COVID-19 vaccination in patients with chronic pulmonary diseases.

INTRODUCTION: Responses to COVID-19 vaccination in patients with chronic pulmonary diseases are poorly characterised. We aimed to describe humoral responses following two doses of BNT162b2 mRNA COVID-19 vaccine and identify risk factors for impaired responses. METHODS: Prospective cohort study including adults with chronic pulmonary diseases and healthcare personnel as controls (1:1). Blood was sampled at inclusion, 3 weeks, 2 and 6 months after first vaccination. We reported antibody concentrations as geometric means with 95% CI of receptor binding domain (RBD)-IgG and neutralising antibody index of inhibition of ACE-2/RBD interaction (%). A low responder was defined as neutralising index in the lowest quartile (primary outcome) or RBD-IgG <225 AU/mL plus neutralising index <25% (secondary outcome), measured at 2 months. We tested associations using Poisson regression. RESULTS: We included 593 patients and 593 controls, 75% of all had neutralising index ≥97% at 2 months. For the primary outcome, 34.7% of patients (n=157/453) and 12.9% of controls (n=46/359) were low responders (p<0.0001). For the secondary outcome, 8.6% of patients (n=39/453) and 1.4% of controls (n=5/359) were low responders (p<0.001). Risk factors associated with low responder included increasing age (per decade, adjusted risk ratio (aRR) 1.17, 95% CI 1.03 to 1.32), Charlson Comorbidity Index (per point) (aRR 1.15, 95% CI 1.05 to 1.26), use of prednisolone (aRR 2.08, 95% CI 1.55 to 2.77) and other immunosuppressives (aRR 2.21, 95% CI 1.65 to 2.97). DISCUSSION: Patients with chronic pulmonary diseases established functional humoral responses to vaccination, however lower than controls. Age, comorbidities and immunosuppression were associated with poor immunological responses.

Modeling of waning immunity after SARS-CoV-2 vaccination and influencing factors.

SARS-CoV-2 vaccines are crucial in controlling COVID-19, but knowledge of which factors determine waning immunity is limited. We examined antibody levels and T-cell gamma-interferon release after two doses of BNT162b2 vaccine or a combination of ChAdOx1-nCoV19 and BNT162b2 vaccines for up to 230 days after the first dose. Generalized mixed models with and without natural cubic splines were used to determine immunity over time. Antibody responses were influenced by natural infection, sex, and age. IgA only became significant in naturally infected. A one-year IgG projection suggested an initial two-phase response in those given the second dose delayed (ChAdOx1/BNT162b2) followed by a more rapid decrease of antibody levels. T-cell responses correlated significantly with IgG antibody responses. Our results indicate that IgG levels will drop at different rates depending on prior infection, age, sex, T-cell response, and the interval between vaccine injections. Only natural infection mounted a significant and lasting IgA response.

Humoral and T-cell response 12 months after the first BNT162b2 vaccination in solid organ transplant recipients and controls: Kinetics, associated factors, and role of SARS-CoV-2 infection.

Introduction: We investigated humoral and T-cell responses within 12 months after first BNT162b2 vaccine in solid organ transplant (SOT) recipients and controls who had received at least three vaccine doses. Furthermore, we compared the immune response in participants with and without previous SARS-CoV-2 infection. Methods: We included adult liver, lung, and kidney transplant recipients, and controls were selected from a parallel cohort of healthcare workers. Results: At 12th-month, the IgG geometric mean concentrations (GMCs) (P<0.001), IgA GMCs (P=0.003), and median IFN-γ (P<0.001) were lower in SOT recipients than in controls. However, in SOT recipients and controls with previous infection, the neutralizing index was 99%, and the IgG, and IgA responses were comparable. After adjustment, female-sex (aOR: 3.6, P<0.009), kidney (aOR: 7.0, P= 0.008) or lung transplantation (aOR: 7.5, P= 0.014), and use of mycophenolate (aOR: 5.2, P=0.03) were associated with low IgG non response. Age (OR:1.4, P=0.038), time from transplantation to first vaccine (OR: 0.45, P<0.035), and previous SARS-CoV-2 infection (OR: 0.14, P<0.001), were associated with low IgA non response. Diabetes (OR:2.4, P=0.044) was associated with T-cell non response. Conclusion: In conclusion, humoral and T-cell responses were inferior in SOT recipients without previous SARS-CoV-2 infection but comparable to controls in SOT recipients with previous infection.

Deep protein representations enable recombinant protein expression prediction.

A crucial process in the production of industrial enzymes is recombinant gene expression, which aims to induce enzyme overexpression of the genes in a host microbe. Current approaches for securing overexpression rely on molecular tools such as adjusting the recombinant expression vector, adjusting cultivation conditions, or performing codon optimizations. However, such strategies are time-consuming, and an alternative strategy would be to select genes for better compatibility with the recombinant host. Several methods for predicting soluble expression are available; however, they are all optimized for the expression host Escherichia coli and do not consider the possibility of an expressed protein not being soluble. We show that these tools are not suited for predicting expression potential in the industrially important host Bacillus subtilis. Instead, we build a B. subtilis-specific machine learning model for expressibility prediction. Given millions of unlabelled proteins and a small labeled dataset, we can successfully train such a predictive model. The unlabeled proteins provide a performance boost relative to using amino acid frequencies of the labeled proteins as input. On average, we obtain a modest performance of 0.64 area-under-the-curve (AUC) and 0.2 Matthews correlation coefficient (MCC). However, we find that this is sufficient for the prioritization of expression candidates for high-throughput studies. Moreover, the predicted class probabilities are correlated with expression levels. A number of features related to protein expression, including base frequencies and solubility, are captured by the model.

Spectrum of Protein Location in Proteomes Captures Evolutionary Relationship Between Species.

The native subcellular location (also referred to as localization or cellular compartment) of a protein is the one in which it acts most frequently; it is one aspect of protein function. Do ten eukaryotic model organisms differ in their location spectrum, i.e., the fraction of its proteome in each of seven major cellular compartments? As experimental annotations of locations remain biased and incomplete, we need prediction methods to answer this question. After systematic bias corrections, the complete but faulty prediction methods appeared to be more appropriate to compare location spectra between species than the incomplete more accurate experimental data. This work compared the location spectra for ten eukaryotes: Homo sapiens (human), Gorilla gorilla (gorilla), Pan troglodytes (chimpanzee), Mus musculus (mouse), Rattus norvegicus (rat), Drosophila melanogaster (fruit/vinegar fly), Anopheles gambiae (African malaria mosquito), Caenorhabitis elegans (nematode), Saccharomyces cerevisiae (baker's yeast), and Schizosaccharomyces pombe (fission yeast). The two largest classes were predicted to be the nucleus and the cytoplasm together accounting for 47-62% of all proteins, while 7-21% of the proteins were predicted in the plasma membrane and 4-15% to be secreted. Overall, the predicted location spectra were largely similar. However, in detail, the differences sufficed to plot trees (UPGMA) and 2D (PCA) maps relating the ten organisms using a simple Euclidean distance in seven states (location classes). The relations based on the simple predicted location spectra captured aspects of cross-species comparisons usually revealed only by much more detailed evolutionary comparisons. Most interestingly, known phylogenetic relations were reproduced better by paralog-only than by ortholog-only trees.

Improved metagenome binning and assembly using deep variational autoencoders.

Despite recent advances in metagenomic binning, reconstruction of microbial species from metagenomics data remains challenging. Here we develop variational autoencoders for metagenomic binning (VAMB), a program that uses deep variational autoencoders to encode sequence coabundance and k-mer distribution information before clustering. We show that a variational autoencoder is able to integrate these two distinct data types without any previous knowledge of the datasets. VAMB outperforms existing state-of-the-art binners, reconstructing 29-98% and 45% more near-complete (NC) genomes on simulated and real data, respectively. Furthermore, VAMB is able to separate closely related strains up to 99.5% average nucleotide identity (ANI), and reconstructed 255 and 91 NC Bacteroides vulgatus and Bacteroides dorei sample-specific genomes as two distinct clusters from a dataset of 1,000 human gut microbiome samples. We use 2,606 NC bins from this dataset to show that species of the human gut microbiome have different geographical distribution patterns. VAMB can be run on standard hardware and is freely available at https://github.com/RasmussenLab/vamb .

Similar genomic patterns of clinical infective endocarditis and oral isolates of Streptococcus sanguinis and Streptococcus gordonii.

Streptococcus gordonii and Streptococcus sanguinis belong to the Mitis group streptococci, which mostly are commensals in the human oral cavity. Though they are oral commensals, they can escape their niche and cause infective endocarditis, a severe infection with high mortality. Several virulence factors important for the development of infective endocarditis have been described in these two species. However, the background for how the commensal bacteria, in some cases, become pathogenic is still not known. To gain a greater understanding of the mechanisms of the pathogenic potential, we performed a comparative analysis of 38 blood culture strains, S. sanguinis (n = 20) and S. gordonii (n = 18) from patients with verified infective endocarditis, along with 21 publicly available oral isolates from healthy individuals, S. sanguinis (n = 12) and S. gordonii (n = 9). Using whole genome sequencing data of the 59 streptococci genomes, functional profiles were constructed, using protein domain predictions based on the translated genes. These functional profiles were used for clustering, phylogenetics and machine learning. A clear separation could be made between the two species. No clear differences between oral isolates and clinical infective endocarditis isolates were found in any of the 675 translated core-genes. Additionally, random forest-based machine learning and clustering of the pan-genome data as well as amino acid variations in the core-genome could not separate the clinical and oral isolates. A total of 151 different virulence genes was identified in the 59 genomes. Among these homologs of genes important for adhesion and evasion of the immune system were found in all of the strains. Based on the functional profiles and virulence gene content of the genomes, we believe that all analysed strains had the ability to become pathogenic.