Social Media Data for Omicron Detection from Unscripted Voice Samples

The success of artificial intelligence in clinical environments relies upon the diversity and availability of training data. In some cases, social media data may be used to counterbalance the limited amount of accessible, well-curated clinical data, but this possibility remains largely unexplored. In this study, we mined YouTube to collect voice data from individuals with self-declared positive COVID-19 tests during time periods in which Omicron was the predominant variant1,2,3, while also sampling non-Omicron COVID-19 variants, other upper respiratory infections (URI), and healthy subjects. The resulting dataset was used to train a DenseNet model to detect the Omicron variant from voice changes. Our model achieved 0.85/0.80 sensitivity/specificity in separating Omicron samples from healthy samples and 0.76/0.70 sensitivity/specificity in separating Omicron samples from symptomatic non-COVID samples. In comparison with past studies, which used scripted voice samples, we showed that leveraging the intra-sample variance inherent to unscripted speech enhanced generalization. Our work introduced novel design paradigms for audio-based diagnostic tools and established the potential of social media data to train digital diagnostic models suitable for real-world deployment.

SARS-CoV-2 infection in Africa: A systematic review and meta-analysis of standardised seroprevalence studies, from January 2020 to December 2021

IntroductionEstimating COVID-19 cumulative incidence in Africa remains problematic due to challenges in contact tracing, routine surveillance systems and laboratory testing capacities and strategies. We undertook a meta-analysis of population-based seroprevalence studies to estimate SARS-CoV-2 seroprevalence in Africa to inform evidence-based decision making on Public Health and Social Measures (PHSM) and vaccine strategy. MethodsWe searched for seroprevalence studies conducted in Africa published 01-01-2020 to 30-12-2021 in Medline, Embase, Web of Science, and Europe PMC (preprints), grey literature, media releases and early results from WHO Unity studies. All studies were screened, extracted, assessed for risk of bias and evaluated for alignment with the WHO Unity protocol for seroepidemiological investigations. We conducted descriptive analyses of seroprevalence and meta-analysed seroprevalence differences by demographic groups, place and time. We estimated the extent of undetected infections by comparing seroprevalence and cumulative incidence of confirmed cases reported to WHO. PROSPERO: CRD42020183634. ResultsWe identified 54 full texts or early results, reporting 151 distinct seroprevalence studies in Africa Of these, 95 (63%) were low/moderate risk of bias studies. SARS-CoV-2 seroprevalence rose from 3.0% [95% CI: 1.0-9.2%] in Q2 2020 to 65.1% [95% CI: 56.3-73.0%] in Q3 2021. The ratios of seroprevalence from infection to cumulative incidence of confirmed cases was large (overall: 97:1, ranging from 10:1 to 958:1) and steady over time. Seroprevalence was highly heterogeneous both within countries - urban vs. rural (lower seroprevalence for rural geographic areas), children vs. adults (children aged 0-9 years had the lowest seroprevalence) - and between countries and African sub-regions (Middle, Western and Eastern Africa associated with higher seroprevalence). ConclusionWe report high seroprevalence in Africa suggesting greater population exposure to SARS-CoV-2 and protection against COVID-19 disease than indicated by surveillance data. As seroprevalence was heterogeneous, targeted PHSM and vaccination strategies need to be tailored to local epidemiological situations.

Global seroprevalence of SARS-CoV-2 antibodies: a systematic review and meta-analysis

BackgroundMany studies report the seroprevalence of severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) antibodies. We aimed to synthesize seroprevalence data to better estimate the level and distribution of SARS-CoV-2 infection, identify high-risk groups, and inform public health decision making. MethodsIn this systematic review and meta-analysis, we searched publication databases, preprint servers, and grey literature sources for seroepidemiological study reports, from January 1, 2020 to December 31, 2020. We included studies that reported a sample size, study date, location, and seroprevalence estimate. We corrected estimates for imperfect test accuracy with Bayesian measurement error models, conducted meta-analysis to identify demographic differences in the prevalence of SARS-CoV-2 antibodies, and meta-regression to identify study-level factors associated with seroprevalence. We compared region-specific seroprevalence data to confirmed cumulative incidence. PROSPERO: CRD42020183634. ResultsWe identified 968 seroprevalence studies including 9.3 million participants in 74 countries. There were 472 studies (49%) at low or moderate risk of bias. Seroprevalence was low in the general population (median 4.5%, IQR 2.4-8.4%); however, it varied widely in specific populations from low (0.6% perinatal) to high (59% persons in assisted living and long-term care facilities). Median seroprevalence also varied by Global Burden of Disease region, from 0.6 % in Southeast Asia, East Asia and Oceania to 19.5% in Sub-Saharan Africa (p<0.001). National studies had lower seroprevalence estimates than regional and local studies (p<0.001). Compared to Caucasian persons, Black persons (prevalence ratio [RR] 3.37, 95% CI 2.64-4.29), Asian persons (RR 2.47, 95% CI 1.96-3.11), Indigenous persons (RR 5.47, 95% CI 1.01-32.6), and multi-racial persons (RR 1.89, 95% CI 1.60-2.24) were more likely to be seropositive. Seroprevalence was higher among people ages 18-64 compared to 65 and over (RR 1.27, 95% CI 1.11-1.45). Health care workers in contact with infected persons had a 2.10 times (95% CI 1.28-3.44) higher risk compared to health care workers without known contact. There was no difference in seroprevalence between sex groups. Seroprevalence estimates from national studies were a median 18.1 times (IQR 5.9-38.7) higher than the corresponding SARS-CoV-2 cumulative incidence, but there was large variation between Global Burden of Disease regions from 6.7 in South Asia to 602.5 in Sub-Saharan Africa. Notable methodological limitations of serosurveys included absent reporting of test information, no statistical correction for demographics or test sensitivity and specificity, use of non-probability sampling and use of non-representative sample frames. DiscussionMost of the population remains susceptible to SARS-CoV-2 infection. Public health measures must be improved to protect disproportionately affected groups, including racial and ethnic minorities, until vaccine-derived herd immunity is achieved. Improvements in serosurvey design and reporting are needed for ongoing monitoring of infection prevalence and the pandemic response. FundingPublic Health Agency of Canada through the COVID-19 Immunity Task Force.

Artificial intelligence driven assessment of routinely collected healthcare data is an effective screening test for COVID-19 in patients presenting to hospital

BackgroundRapid identification of COVID-19 is important for delivering care expediently and maintaining infection control. The early clinical course of SARS-CoV-2 infection can be difficult to distinguish from other undifferentiated medical presentations to hospital, however for operational reasons SARS-CoV-2 PCR testing can take up to 48 hours. Artificial Intelligence (AI) methods, trained using routinely collected clinical data, may allow front-door screening for COVID-19 within the first hour of presentation. MethodsDemographic, routine and prior clinical data were extracted for 170,510 sequential presentations to emergency and acute medical departments at a large UK teaching hospital group. We applied multivariate logistic regression, random forests and extreme gradient boosted trees to distinguish emergency department (ED) presentations and admissions due to COVID-19 from pre-pandemic controls. We performed stepwise addition of clinical feature sets and assessed performance using stratified 10-fold cross validation. Models were calibrated during training to achieve sensitivities of 70, 80 and 90% for identifying patients with COVID-19. To simulate real-world performance at different stages of an epidemic, we generated test sets with varying prevalences of COVID-19 and assessed predictive values. We prospectively validated our models for all patients presenting or admitted to our hospital group between 20th April and 6th May 2020, comparing model predictions to PCR test results. ResultsPresentation laboratory blood tests, point of care blood gas, and vital signs measurements for 115,394 emergency presentations and 72,310 admissions were analysed. Presentation laboratory tests and vital signs were most predictive of COVID-19 (maximum area under ROC curve [AUROC] 0.904 and 0.823, respectively). Sequential addition of informative variables improved model performance to AUROC 0.942. We developed two early-detection models to identify COVID-19, achieving sensitivities and specificities of 77.4% and 95.7% for our ED model amongst patients attending hospital, and 77.4% and 94.8% for our Admissions model amongst patients being admitted. Both models offer high negative predictive values (>99%) across a range of prevalences (<5%). In a two-week prospective validation period, our ED and Admissions models demonstrated 92.3% and 92.5% accuracy (AUROC 0.881 and 0.871 respectively) for all patients presenting or admitted to a large UK teaching hospital group. A sensitivity analysis to account for uncertainty in negative PCR results improves apparent accuracy (95.1% and 94.1%) and NPV (99.0% and 98.5%). Three laboratory blood markers, Eosinophils, Basophils, and C-Reactive Protein, alongside Calcium measured on blood-gas, and presentation Oxygen requirement were the most informative variables in our models. ConclusionArtificial intelligence techniques perform effectively as a screening test for COVID-19 in emergency departments and hospital admission units. Our models support rapid exclusion of the illness using routinely collected and readily available clinical measurements, guiding streaming of patients during the early phase of admission. BriefThe early clinical course of SARS-CoV-2 infection can be difficult to distinguish from other undifferentiated medical presentations to hospital, however viral specific real-time polymerase chain reaction (RT-PCR) testing has limited sensitivity and can take up to 48 hours for operational reasons. In this study, we develop two early-detection models to identify COVID-19 using routinely collected data typically available within one hour (laboratory tests, blood gas and vital signs) during 115,394 emergency presentations and 72,310 admissions to hospital. Our emergency department (ED) model achieved 77.4% sensitivity and 95.7% specificity (AUROC 0.939) for COVID-19 amongst all patients attending hospital, and Admissions model achieved 77.4% sensitivity and 94.8% specificity (AUROC 0.940) for the subset admitted to hospital. Both models achieve high negative predictive values (>99%) across a range of prevalences (<5%), facilitating rapid exclusion during triage to guide infection control. We prospectively validated our models across all patients presenting and admitted to a large UK teaching hospital group in a two-week test period, achieving 92.3% (n= 3,326, NPV: 97.6%, AUROC: 0.881) and 92.5% accuracy (n=1,715, NPV: 97.7%, AUROC: 0.871) in comparison to RT-PCR results. Sensitivity analyses to account for uncertainty in negative PCR results improves apparent accuracy (95.1% and 94.1%) and NPV (99.0% and 98.5%). Our artificial intelligence models perform effectively as a screening test for COVID-19 in emergency departments and hospital admission units, offering high impact in settings where rapid testing is unavailable.

Cadaveric sperm induces intergeneric androgenesis in the fish, Hemigrammus caudovittatus.

Intergeneric androgenetic golden Buenos Aires tetra (BT), Hemigrammus caudovittatus was generated using sperm drawn from post-mortem males preserved at -20 degrees C for 10, 20, 30 and 40 days or fresh sperm to activate the UV-irradiated oocytes of black widow tetra (WT), Gymnocorymbus ternetzi. UV-irradiation (4.2 W/m(2)) of the oocytes for 3 min inactivated their nuclear genome. Fry hatched out from these activated oocytes were haploids; suffering haploid syndrome, they died before or within 48 h after hatching. Fresh BT sperm activated 95% oocytes; however, the sperm drawn from post-mortem males preserved at -20 degrees C for 60 (within glycerol packing) and 30 days (without glycerol packing) activated only 24 and 19% oocytes, respectively. Following activation, diploidy was restored by shocking the 25-min-old embryos at 41 degrees C for 2 min. Nuclear genomic inactivation of the oocytes was confirmed by (i) production of 100% haploids, (ii) karyotype and erythrocyte measurements, (iii) phenotypic markers, (iv) progeny testing and (v) species-specific marker. At hatching, survival of androgenotes decreased from 11% for those induced with fresh sperm to 4% for those generated using sperm from 30-day-old post-mortem males. Reproductive performance of the 'fresh' and 'cadaveric' F(0) and F(1) androgenetic males (Y(2)Y(2)) was superior to the control (X(1)Y(2)). Crosses involving homozygous (Y(2)Y(2)) 'fresh' F(0) androgenetic males with heterozygous females (X(1)X(2)) and F(0) homozygous males (Y(2)Y(2)) with females (X(2)X(2)) produced 2-4% unexpected female progenies. Paternal autosomes, inherited by the homozygous androgenetic female (X(2)X(2)), induced the production of female progenies in significantly less number of crosses than the crosses with heterozygous females (X(1)X(2)), which carried equal number of paternal and maternal autosomes. PCR analyses of the genomic DNA of normal male and unexpected F(1) and F(2) female progenies amplified by DMRT 1 specific primer produced bands of 237 and 300 bp length, and thereby confirmed that these unexpected females were genetic males. RAPD analyses of the androgenetic progenies showed that their genome was not contaminated with maternal genome.

GFP reporter gene confirms paternity in the androgenote Buenos Aires tetra, Hemigrammus caudovittatus.

A protocol for successful induction of androgenetic cloning of the Buenos Aires tetra (BT), Hemigrammus caudovittatus, with contrasting gray and golden strains is described. At the intensity of 4.2 W/m(2), UV irradiation for 2.75 min totally inactivated the maternal genome in eggs of gray BT. Following activation by sperm of golden BT, the 25-min-old embryos were shocked at 41 degrees C for 2 min to restore diploidy. Interestingly, the hatching success of the haploid fry was always higher than that of the diploid fry, indicating that the enhanced homozygosity (Y(2)Y(2)) is more deleterious than haploidy. Maternal genomic inactivation was confirmed by (i) expression of green fluorescent protein (GFP) gene in the 6-16 hr old live haploid and aneuploid embryos, (ii) golden body color in the diploid fry and adult and (iii) progeny testing. Survival of androgenotes was 10% at hatching and 6% at sexual maturity. Reproductive performance of F(0) and F(1) males (Y(2)Y(2)) was superior to that of normal ones (X(1)Y(2)), but that of the F(0) and F(1) females (X(2)X(2)) was inferior to the control (X(1)X(2)). Of 21 crosses involving homozygous androgenetic (Y(2)Y(2)) males and heterozygous (X(1)X(2)) females, 7 of them (33%) produced 3-9% unexpected female progenies. But only a single cross (14%) generated 3-4% unexpected female progenies, when 7 pairs of homozygous androgenetic (Y(2)Y(2)) males and (X(2)X(2)) females were crossed. Hence, the paternal autosomes, inherited by the homozygous androgenetic female (X(2)X(2)), produced female progenies in significantly less number of crosses, also at lower frequencies than the crosses with heterozygous females (X(1)X(2)), which carried an equal number of paternal and maternal autosomes. However, progenies resulting from the cross between gray female (X(1)X(2)) and golden male (Y(2)Y(2)), after undergoing androgenesis, were males, with paternal chromosomes alone, indicating that the presence of Y(2)Y(2) appears to override the modifying effect of autosomes, but the paternal or maternal autosomes seemed to override the single Y(2) present with X(1) or X(2), and induced the production of unexpected female progenies. Using Double sex Mab3 related transcription factor (DMRT 1)-specific primers, PCR analyses of the genomic DNA of the normal (X(1)Y(2)) and androgenetic males (X(1)Y(2)) produced two amplicons of 237 and 300 bp length. However, they were not detectable in the female (X(1)X(2)) genomic DNA, which amplified only one amplicon of 100 bp. Genomic DNA extracted from the 18 unexpected female progenies expressed the (X(1)Y(2)) genotype-specific banding pattern with two amplicons of 237 and 300 bp length and thereby confirmed that they were genotypic males. A partial sequencing of the male-specific sequence indicated that DMRT 1-specific primer was bound to the fragment of the genomic DNA of the male tetra, although the male-specific sequence of DMRT 1 was not completely detectable.