A catalytically active oscillator made from small organic molecules.

Oscillatory systems regulate many biological processes, including key cellular functions such as metabolism and cell division, as well as larger-scale processes such as circadian rhythm and heartbeat1-4. Abiotic chemical oscillations, discovered originally in inorganic systems5,6, inspired the development of various synthetic oscillators for application as autonomous time-keeping systems in analytical chemistry, materials chemistry and the biomedical field7-17. Expanding their role beyond that of a pacemaker by having synthetic chemical oscillators periodically drive a secondary function would turn them into significantly more powerful tools. However, this is not trivial because the participation of components of the oscillator in the secondary function might jeopardize its time-keeping ability. We now report a small molecule oscillator that can catalyse an independent chemical reaction in situ without impairing its oscillating properties. In a flow system, the concentration of the catalytically active product of the oscillator shows sustained oscillations and the catalysed reaction is accelerated only during concentration peaks. Augmentation of synthetic oscillators with periodic catalytic action allows the construction of complex systems that, in the future, may benefit applications in automated synthesis, systems and polymerization chemistry and periodic drug delivery.

EF1α-associated protein complexes affect dendritic spine plasticity by regulating microglial phagocytosis in Fmr1 knock-out mice.

Fragile X syndrome (FXS) is the most common inherited cause of intellectual disability. There is no specific treatment for FXS due to the lack of therapeutic targets. We report here that Elongation Factor 1α (EF1α) forms a complex with two other proteins: Tripartite motif-containing protein 3 (TRIM3) and Murine double minute (Mdm2). Both EF1α-Mdm2 and EF1α-TRIM3 protein complexes are increased in the brain of Fmr1 knockout mice as a result of FMRP deficiency, which releases the normal translational suppression of EF1α mRNA and increases EF1α protein levels. Increased EF1α-Mdm2 complex decreases PSD-95 ubiquitination (Ub-PSD-95) and Ub-PSD-95-C1q interaction. The elevated level of TRIM3-EF1α complex is associated with decreased TRIM3-Complement Component 3 (C3) complex that inhibits the activation of C3. Both protein complexes thereby contribute to a reduction in microglia-mediated phagocytosis and dendritic spine pruning. Finally, we created a peptide that disrupts both protein complexes and restores dendritic spine plasticity and behavioural deficits in Fmr1 knockout mice. The EF1α-Mdm2 and EF1α-TRIM3 complexes could thus be new therapeutic targets for FXS.

The potential of federated learning for public health purposes: a qualitative analysis of GDPR compliance, Europe, 2021.

BackgroundThe wide application of machine learning (ML) holds great potential to improve public health by supporting data analysis informing policy and practice. Its application, however, is often hampered by data fragmentation across organisations and strict regulation by the General Data Protection Regulation (GDPR). Federated learning (FL), as a decentralised approach to ML, has received considerable interest as a means to overcome the fragmentation of data, but it is yet unclear to which extent this approach complies with the GDPR.AimOur aim was to understand the potential data protection implications of the use of federated learning for public health purposes.MethodsBuilding upon semi-structured interviews (n = 14) and a panel discussion (n = 5) with key opinion leaders in Europe, including both FL and GDPR experts, we explored how GDPR principles would apply to the implementation of FL within public health.ResultsWhereas this study found that FL offers substantial benefits such as data minimisation, storage limitation and effective mitigation of many of the privacy risks of sharing personal data, it also identified various challenges. These challenges mostly relate to the increased difficulty of checking data at the source and the limited understanding of potential adverse outcomes of the technology.ConclusionSince FL is still in its early phase and under rapid development, it is expected that knowledge on its impracticalities will increase rapidly, potentially addressing remaining challenges. In the meantime, this study reflects on the potential of FL to align with data protection objectives and offers guidance on GDPR compliance.

Prevalence and Severity of Symptoms 3 Months After Infection With SARS-CoV-2 Compared to Test-Negative and Population Controls in the Netherlands.

BACKGROUND: This prospective study assesses symptoms 3 months after severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) infection compared to test-negative and population controls, and the effect of vaccination prior to infection. METHODS: Participants enrolled after a positive (cases) or negative (test-negative controls) SARS-CoV-2 test, or after invitation from the general population (population controls). After 3 months, participants indicated presence of 41 symptoms and severity of 4 symptoms. Permutation tests were used to select symptoms significantly elevated in cases compared to controls and to compare symptoms between cases that were vaccinated or unvaccinated prior to infection. RESULTS: In total, 9166 cases, 1698 symptomatic but test-negative controls, and 3708 population controls enrolled. At 3 months, 13 symptoms, and severity of fatigue, cognitive impairment, and dyspnea were significantly elevated incases compared to controls. Of cases, 48.5% reported ≥1 significantly elevated symptom compared to 29.8% of test-negative controls and 26.0% of population controls. Effect of vaccination could be determined for cases aged <65 years, and was significantly protective for loss of smell and taste but not for other symptoms. DISCUSSION: Three months after SARS-CoV-2 infection, almost half of cases report symptoms, which was higher than background prevalence and test-negative prevalence. Vaccination prior to infection was protective against loss of smell and taste in cases aged <65 years.

The Bigger Fish: A Comparison of Meta-Learning QSAR Models on Low-Resourced Aquatic Toxicity Regression Tasks.

Toxicological information as needed for risk assessments of chemical compounds is often sparse. Unfortunately, gathering new toxicological information experimentally often involves animal testing. Simulated alternatives, e.g., quantitative structure-activity relationship (QSAR) models, are preferred to infer the toxicity of new compounds. Aquatic toxicity data collections consist of many related tasks─each predicting the toxicity of new compounds on a given species. Since many of these tasks are inherently low-resource, i.e., involve few associated compounds, this is challenging. Meta-learning is a subfield of artificial intelligence that can lead to more accurate models by enabling the utilization of information across tasks. In our work, we benchmark various state-of-the-art meta-learning techniques for building QSAR models, focusing on knowledge sharing between species. Specifically, we employ and compare transformational machine learning, model-agnostic meta-learning, fine-tuning, and multi-task models. Our experiments show that established knowledge-sharing techniques outperform single-task approaches. We recommend the use of multi-task random forest models for aquatic toxicity modeling, which matched or exceeded the performance of other approaches and robustly produced good results in the low-resource settings we studied. This model functions on a species level, predicting toxicity for multiple species across various phyla, with flexible exposure duration and on a large chemical applicability domain.

Predictors of healthy physiological aging across generations in a 30-year population-based cohort study: the Doetinchem Cohort Study.

BACKGROUND: Predicting healthy physiological aging is of major interest within public health research. However, longitudinal studies into predictors of healthy physiological aging that include numerous exposures from different domains (i.e. the exposome) are scarce. Our aim is to identify the most important exposome-related predictors of healthy physiological aging over the life course and across generations. METHODS: Data were used from 2815 participants from four generations (generation 1960s/1950s/1940s/1930s aged respectively 20-29/30-39/40-49/50-59 years old at baseline, wave 1) of the Doetinchem Cohort Study who were measured every 5 years for 30 years. The Healthy Aging Index, a physiological aging index consisting of blood pressure, glucose, creatinine, lung function, and cognitive functioning, was measured at age 46-85 years (wave 6). The average exposure and trend of exposure over time of demographic, lifestyle, environmental, and biological exposures were included, resulting in 86 exposures. Random forest was used to identify important predictors. RESULTS: The most important predictors of healthy physiological aging were overweight-related (BMI, waist circumference, waist/hip ratio) and cholesterol-related (using cholesterol lowering medication, HDL and total cholesterol) measures. Diet and educational level also ranked in the top of important exposures. No substantial differences were observed in the predictors of healthy physiological aging across generations. The final prediction model's performance was modest with an R2 of 17%. CONCLUSIONS: Taken together, our findings suggest that longitudinal cardiometabolic exposures (i.e. overweight- and cholesterol-related measures) are most important in predicting healthy physiological aging. This finding was similar across generations. More work is needed to confirm our findings in other study populations.

Predicting self-perceived general health status using machine learning: an external exposome study.

BACKGROUND: Self-perceived general health (SPGH) is a general health indicator commonly used in epidemiological research and is associated with a wide range of exposures from different domains. However, most studies on SPGH only investigated a limited set of exposures and did not take the entire external exposome into account. We aimed to develop predictive models for SPGH based on exposome datasets using machine learning techniques and identify the most important predictors of poor SPGH status. METHODS: Random forest (RF) was used on two datasets based on personal characteristics from the 2012 and 2016 editions of the Dutch national health survey, enriched with environmental and neighborhood characteristics. Model performance was determined using the area under the curve (AUC) score. The most important predictors were identified using a variable importance procedure and individual effects of exposures using partial dependence and accumulated local effect plots. The final 2012 dataset contained information on 199,840 individuals and 81 variables, whereas the final 2016 dataset had 244,557 individuals with 91 variables. RESULTS: Our RF models had overall good predictive performance (2012: AUC = 0.864 (CI: 0.852-0.876); 2016: AUC = 0.890 (CI: 0.883-0.896)) and the most important predictors were "Control of own life", "Physical activity", "Loneliness" and "Making ends meet". Subjects who felt insufficiently in control of their own life, scored high on the De Jong-Gierveld loneliness scale or had difficulty in making ends meet were more likely to have poor SPGH status, whereas increased physical activity per week reduced the probability of poor SPGH. We observed associations between some neighborhood and environmental characteristics, but these variables did not contribute to the overall predictive strength of the models. CONCLUSIONS: This study identified that within an external exposome dataset, the most important predictors for SPGH status are related to mental wellbeing, physical exercise, loneliness, and financial status.

Introduction of a newly created AW stereotactic frame: a phantom-based accuracy evaluation and an initial experience in clinical usage.

BACKGROUND: A new stereotactic frame was created in 2015, based on a linear algorithm. It is called Albert Wong (AW) frame. A simple AW stereo-calculator was also designed based on Excel® (Microscoft Corporation, Redmond, WA) programme for the frame. OBJECTIVE: The aim of this study is to test the accuracy of the AW frame by a direct head to head comparison with CRW® frame (Integra Life Sciences, Plainsboro, NJ) on a phantom. METHODS: This is a prospective pilot cross-sectional phantom study with a total of 42 (21 for AW and 21 for CRW®) laboratory testings performed in 2017 at our institute to compare the accuracies of both frames in a consecutive manner. A phantom (BL phantom) was newly created, where targets can be placed at different heights and positions on a platform attached under the frame for accuracy testing comparing between the AW and CRW® frames. RESULTS: A comparable accuracy testing results were observed between the AW and CRW® frames of 0.64 mm versus 1.07 mm respectively. Approval from the local ethics committee for a clinical trial was obtained. We report on three case illustrations who had the AW frame-based biopsies with definitive diagnoses and without any post-biopsy related complication. CONCLUSION: AW frame successfully demonstrated a good accuracy of 0.64 mm in phantom testing using the BL phantom by a linear algorithmic calculation. The clinical trial with three patients demonstrated definitive diagnoses and safety with its use.

History Dependence in a Chemical Reaction Network Enables Dynamic Switching.

This work describes an enzymatic autocatalytic network capable of dynamic switching under out-of-equilibrium conditions. The network, wherein a molecular fuel (trypsinogen) and an inhibitor (soybean trypsin inhibitor) compete for a catalyst (trypsin), is kept from reaching equilibria using a continuous flow stirred tank reactor. A so-called 'linear inhibition sweep' is developed (i.e., a molecular analogue of linear sweep voltammetry) to intentionally perturb the competition between autocatalysis and inhibition, and used to demonstrate that a simple molecular system, comprising only three components, is already capable of a variety of essential neuromorphic behaviors (hysteresis, synchronization, resonance, and adaptation). This research provides the first steps in the development of a strategy that uses the principles in systems chemistry to transform chemical reaction networks into platforms capable of neural network computing.

Genetic modification of PIN genes induces causal mechanisms of stay-green drought adaptation phenotype.

The stay-green trait is recognized as a key drought adaptation mechanism in cereals worldwide. Stay-green sorghum plants exhibit delayed senescence of leaves and stems, leading to prolonged growth, a reduced risk of lodging, and higher grain yield under end-of-season drought stress. More than 45 quantitative trait loci (QTL) associated with stay-green have been identified, including two major QTL (Stg1 and Stg2). However, the contributing genes that regulate functional stay-green are not known. Here we show that the PIN FORMED family of auxin efflux carrier genes induce some of the causal mechanisms driving the stay-green phenotype in sorghum, with SbPIN4 and SbPIN2 located in Stg1 and Stg2, respectively. We found that nine of 11 sorghum PIN genes aligned with known stay-green QTL. In transgenic studies, we demonstrated that PIN genes located within the Stg1 (SbPIN4), Stg2 (SbPIN2), and Stg3b (SbPIN1) QTL regions acted pleiotropically to modulate canopy development, root architecture, and panicle growth in sorghum, with SbPIN1, SbPIN2, and SbPIN4 differentially expressed in various organs relative to the non-stay-green control. The emergent consequence of such modifications in canopy and root architecture is a stay-green phenotype. Crop simulation modelling shows that the SbPIN2 phenotype can increase grain yield under drought.

The D2R-DISC1 protein complex and associated proteins are altered in schizophrenia and normalized with antipsychotic treatment.

BACKGROUND: For decades, the dopamine D2 receptor (D2R) has been known as the main target of antipsychotic medications, but the mechanism for antipsychotic effects beyond this pharmacological target remains unclear. Disrupted-in-schizophrenia 1 (DISC1) is a gene implicated in the etiology of schizophrenia, and we have found elevated levels of the D2R-DISC1 complex in the postmortem brain tissue of patients with schizophrenia. METHODS: We used coimmunoprecipitation to measure D2R-DISC1 complex levels in peripheral blood samples from patients with schizophrenia and unaffected controls in 3 cohorts (including males and females) from different hospitals. We also used label-free mass spectrometry to conduct proteomic analysis of these samples. RESULTS: Levels of the D2R-DISC1 complex were elevated in the peripheral blood samples of patients with schizophrenia from 3 independent cohorts, and were normalized with antipsychotic treatment. Proteomic analysis of the blood samples from patients with high D2R-DISC1 complex levels that were normalized with antipsychotic treatment revealed a number of altered proteins and pathways associated with D2R, DISC1 and the D2R-DISC1 complex. We identified additional proteins and pathways that were associated with antipsychotic treatment in schizophrenia, and that may also be novel targets for schizophrenia treatment. LIMITATIONS: Sample sizes were relatively small, but were sufficient to detect associations between D2R-DISC1 levels, schizophrenia and treatment response. The relevance of leukocyte changes to the symptoms of schizophrenia is unknown. The coimmunoprecipitation lanes included several nonspecific bands. CONCLUSION: Levels of the D2R-DISC1 complex were elevated in patients with schizophrenia and reduced with antipsychotic treatment. This finding reinforces the independent role of each protein in schizophrenia. Our results enhanced our understanding of the molecular pathways involved in schizophrenia and in antipsychotic medications, and identified novel potential molecular targets for treating schizophrenia.

Factors affecting diagnostic yield in stereotactic biopsy for brain lesions: a 5-year single-center series.

The objective of this study is to determine the factors that are associated with the diagnostic yield of stereotactic brain biopsy. A retrospective analysis was performed on 50 consecutive patients who underwent stereotactic brain biopsies in a single institute from 2014 to 2019. Variables including age, gender, lesion topography and characteristics, biopsy methods, and surgeon's experience were analyzed along with diagnostic rate. This study included 31 male and 19 female patients with a mean age of 48.4 (range: 1-76). Of these, 25 underwent frameless brain-suite stereotactic biopsies, 15 were frameless Portable Brain-lab® stereotactic biopsies and 10 were frame-based CRW® stereotactic biopsies. There was no statistical difference between the diagnostic yield of the three methods. The diagnostic yield in our series was 76%. Age, gender, and biopsy methods had no impact on diagnostic yield. Periventricular and pineal lesion biopsies were significantly associated with negative diagnostic yield (p = 0.01) whereas larger lesions were significantly associated with a positive yield (p = 0.01) with the mean volume of lesions in the positive yield group (13.6 cc) being higher than the negative yield group (7 cc). The diagnostic yields seen between senior and junior neurosurgeons in the biopsy procedure were 95% and 63%, respectively (p = 0.02). Anatomical location of the lesion, volume of the lesion, and experience of the surgeon have significant impacts on the diagnostic yield in stereotactic brain biopsy. There was no statistical difference between the diagnostic yield of the three methods, age, gender, and depth of lesion.

Hotter, drier, CRISPR: the latest edit on climate change.

KEY MESSAGE: Integrating CRISPR/Cas9 genome editing into modern breeding programs for crop improvement in cereals. Global climate trends in many agricultural regions have been rapidly changing over the past decades, and major advances in global food systems are required to ensure food security in the face of these emerging challenges. With increasing climate instability due to warmer temperatures and rising CO2 levels, the productivity of global agriculture will continue to be negatively impacted. To combat these growing concerns, creative approaches will be required, utilising all the tools available to produce more robust and tolerant crops with increased quality and yields under more extreme conditions. The integration of genome editing and transgenics into current breeding strategies is one promising solution to accelerate genetic gains through targeted genetic modifications, producing crops that can overcome the shifting climate realities. This review focuses on how revolutionary genome editing tools can be directly implemented into breeding programs for cereal crop improvement to rapidly counteract many of the issues affecting agriculture production in the years to come.

Socioeconomic differences in healthcare expenditure and utilization in The Netherlands.

BACKGROUND: Worldwide, socioeconomic differences in health and use of healthcare resources have been reported, even in countries providing universal healthcare coverage. However, it is unclear how large these socioeconomic differences are for different types of care and to what extent health status plays a role. Therefore, our aim was to examine to what extent healthcare expenditure and utilization differ according to educational level and income, and whether these differences can be explained by health inequalities. METHODS: Data from 18,936 participants aged 25-79 years of the Dutch Health Interview Survey were linked at the individual level to nationwide claims data that included healthcare expenditure covered in 2017. For healthcare utilization, participants reported use of different types of healthcare in the past 12 months. The association of education/income with healthcare expenditure/utilization was studied separately for different types of healthcare such as GP and hospital care. Subsequently, analyses were adjusted for general health, physical limitations, and mental health. RESULTS: For most types of healthcare, participants with lower educational and income levels had higher healthcare expenditure and used more healthcare compared to participants with the highest educational and income levels. Total healthcare expenditure was approximately between 50 and 150 % higher (depending on age group) among people in the lowest educational and income levels. These differences generally disappeared or decreased after including health covariates in the analyses. After adjustment for health, socioeconomic differences in total healthcare expenditure were reduced by 74-91 %. CONCLUSIONS: In this study among Dutch adults, lower socioeconomic status was associated with increased healthcare expenditure and utilization. These socioeconomic differences largely disappeared after taking into account health status, which implies that, within the universal Dutch healthcare system, resources are being spent where they are most needed. Improving health among lower socioeconomic groups may contribute to decreasing health inequalities and healthcare spending.

Clinical Outcome of Tuberculous Meningitis with Hydrocephalus - A Retrospective Study.

BACKGROUND: To study the clinical outcome of tuberculous meningitis with hydrocephalus (TBMH) and the factors contributing to its poor clinical outcome. METHODS: Clinical data of 143 adult patients diagnosed with TBM over a 6-year period in two tertiary hospitals in Malaysia were retrospectively reviewed. Relevant clinical and radiological data were studied. Patients with TBMH were further analysed based on their clinical grade and rendered treatment to identify associated factors and outcome of this subgroup of patients. The functional outcome of patients was assessed at 12 months from treatment. RESULTS: The mean age of patients was 35.6 (12.4) years old, with a male gender predominance of 67.1%. Forty-four percent had TBMH, of which 42.9% had surgical intervention. In the good modified Vellore grade, 76.5% was managed medically with concurrent anti-tuberculosis treatment (ATT), steroids and osmotic agents. Four patients had surgery early in the disease as they did not respond to medical therapy and reported a good outcome subsequently. Poor outcome (65.2%) was seen in the poor modified Vellore grade despite medical and surgical intervention. Multivariate model multiple Cox regression showed significant results for seizure (adjusted hazard ratio [aHR]: 15.05; 95% CI: 3.73, 60.78), Glasgow coma scale (GCS) (aHR: 0.79; 95% CI: 0.70, 0.89) and cerebrospinal fluid (CSF) cell count (aHR: 1.11; 95% CI: 1.05, 1.17). CONCLUSION: Hydrocephalus was seen in 44% of patients in this study. GCS score, seizure and high CSF cell count were factors associated with a poor prognosis in TBM. Patients with TBMH treated medically (TBMHM) had better survival function compared to TBMH patients undergoing surgical intervention (TBMHS) (P-value < 0.001). This retrospective study emphasises that TBMH is still a serious illness as 47.6% of the patients had poor outcome despite adequate treatment.

History of Neurosurgery in Malaysia: The Past, Present and Future.

The history of neurosurgery in Malaysia traces back to 1962 and is filled with stories of vibrant and humble neurosurgeons who have dedicated their life to patients and professions. The early development of neurological and neurosurgical services begins from the establishment of the neurosurgery unit at Hospital Kuala Lumpur (HKL), followed by the foundation of the Tunku Abdul Rahman Neuroscience Institute (IKTAR). Due to the exponentially increased demand for the care of neurosurgical patients, many universities and government hospitals have opened their neurosurgical units. In 2001, the formal residency training programme (USM Masters in Neurosurgery) started and since then has produced qualified neurosurgeons that empowered and shaped the present generation. The formation of the Neurosurgical Association of Malaysia (NAM) is another turning point towards bidirectional collaboration with the World Federation of Neurosurgical Societies (WFNS). Many opportunities were created for educational activities and the expansion of subspecialties in neurosurgery. This article describes the impact of the past neurosurgeons and the endeavors that they had gone through; the present neurosurgeons who pioneered the current neurosurgical services in Malaysia, and the future neurosurgeons that will continue the legacy and bring neurosurgery further ahead in this country.

Meningococcal W135 Disease Vaccination Intent, the Netherlands, 2018-2019.

To control the rise in Neisseria meningitidis strain W infections, during 2018-2019, the Netherlands launched a catch-up meningococcal conjugate (MenACWY) vaccination campaign for teenagers (13-18 years of age). Applying a mental models approach, we surveyed teenagers and their parents about their knowledge and beliefs about meningococcal disease, the MenACWY vaccination, vaccinations in general, and their MenACWY vaccination intentions. Using random forest analysis, we studied predictions of vaccination intentions by knowledge and beliefs. Survey response rate was 52.8% among teenagers and 59.4% among parents. MenACWY vaccination intentions were best predicted by knowledge and beliefs about vaccinations in general, surpassing knowledge and beliefs about meningococcal disease and the MenACWY vaccination. For teenagers, their parents' intention that the teenager be vaccinated was a strong predictor of the teenagers' own vaccination intention. To optimize vaccination uptake during future outbreaks, we recommend that communications emphasize the effectiveness and safety of vaccines and continue to focus on parents.

Dynamic Environments as a Tool to Preserve Desired Output in a Chemical Reaction Network.

Current efforts to design functional molecular systems have overlooked the importance of coupling out-of-equilibrium behaviour with changes in the environment. Here, the authors use an oscillating reaction network and demonstrate that the application of environmental forcing, in the form of periodic changes in temperature and in the inflow of the concentration of one of the network components, removes the dependency of the periodicity of this network on temperature or flow rates and enforces a stable periodicity across a wide range of conditions. Coupling a system to a dynamic environment can thus be used as a simple tool to regulate the output of a network. In addition, the authors show that coupling can also induce an increase in behavioural complexity to include quasi-periodic oscillations.

Clustering of chronic hepatitis B screening intentions in social networks of Moroccan immigrants in the Netherlands.

BACKGROUND: Early detection, identification, and treatment of chronic hepatitis B through screening is vital for those at increased risk, e.g. born in hepatitis B endemic countries. In the Netherlands, Moroccan immigrants show low participation rates in health-related screening programmes. Since social networks influence health behaviour, we investigated whether similar screening intentions for chronic hepatitis B cluster within social networks of Moroccan immigrants. METHODS: We used respondent-driven sampling (RDS) where each participant ("recruiter") was asked to complete a questionnaire and to recruit three Moroccans ("recruitees") from their social network. Logistic regression analyses were used to analyse whether the recruiters' intention to request a screening test was similar to the intention of their recruitees. RESULTS: We sampled 354 recruiter-recruitee pairs: for 154 pairs both participants had a positive screening intention, for 68 pairs both had a negative screening intention, and the remaining 132 pairs had a discordant intention to request a screening test. A tie between a recruiter and recruitee was associated with having the same screening intention, after correction for sociodemographic variables (OR 1.70 [1.15-2.51]). CONCLUSIONS: The findings of our pilot study show clustering of screening intention among individuals in the same network. This provides opportunities for social network interventions to encourage participation in hepatitis B screening initiatives.

Monitoring stance towards vaccination in twitter messages.

BACKGROUND: We developed a system to automatically classify stance towards vaccination in Twitter messages, with a focus on messages with a negative stance. Such a system makes it possible to monitor the ongoing stream of messages on social media, offering actionable insights into public hesitance with respect to vaccination. At the moment, such monitoring is done by means of regular sentiment analysis with a poor performance on detecting negative stance towards vaccination. For Dutch Twitter messages that mention vaccination-related key terms, we annotated their stance and feeling in relation to vaccination (provided that they referred to this topic). Subsequently, we used these coded data to train and test different machine learning set-ups. With the aim to best identify messages with a negative stance towards vaccination, we compared set-ups at an increasing dataset size and decreasing reliability, at an increasing number of categories to distinguish, and with different classification algorithms. RESULTS: We found that Support Vector Machines trained on a combination of strictly and laxly labeled data with a more fine-grained labeling yielded the best result, at an F1-score of 0.36 and an Area under the ROC curve of 0.66, considerably outperforming the currently used sentiment analysis that yielded an F1-score of 0.25 and an Area under the ROC curve of 0.57. We also show that the recall of our system could be optimized to 0.60 at little loss of precision. CONCLUSION: The outcomes of our study indicate that stance prediction by a computerized system only is a challenging task. Nonetheless, the model showed sufficient recall on identifying negative tweets so as to reduce the manual effort of reviewing messages. Our analysis of the data and behavior of our system suggests that an approach is needed in which the use of a larger training dataset is combined with a setting in which a human-in-the-loop provides the system with feedback on its predictions.