Inspiration

It's hard to read and understand scientific literature, even for undergraduate students in life sciences/computer sciences. We wanted to create a quick way for people absorb new research relevant to the ongoing pandemic.

What it does

Provides a summary of all recent COVID-19 publications around the world and filters them by keywords associated with the publication type, that are of relevance to the general public. Also shows the trajectory of publication keywords over time in an interactive graph.

Examples of summarizer:

(1) ORIGINAL:

'BACKGROUND: In December, 2019, a pneumonia associated with the 2019 novel coronavirus (2019-nCoV) emerged in Wuhan, China. We aimed to further clarify the epidemiological and clinical characteristics of 2019-nCoV pneumonia.\nMETHODS: In this retrospective, single-centre study, we included all confirmed cases of 2019-nCoV in Wuhan Jinyintan Hospital from Jan 1 to Jan 20, 2020. Cases were confirmed by real-time RT-PCR and were analysed for epidemiological, demographic, clinical, and radiological features and laboratory data. Outcomes were followed up until Jan 25, 2020.\nFINDINGS: Of the 99 patients with 2019-nCoV pneumonia, 49 (49%) had a history of exposure to the Huanan seafood market. The average age of the patients was 55·5 years (SD 13·1), including 67 men and 32 women. 2019-nCoV was detected in all patients by real-time RT-PCR. 50 (51%) patients had chronic diseases. Patients had clinical manifestations of fever (82 [83%] patients), cough (81 [82%] patients), shortness of breath (31 [31%] patients), muscle ache (11 [11%] patients), confusion (nine [9%] patients), headache (eight [8%] patients), sore throat (five [5%] patients), rhinorrhoea (four [4%] patients), chest pain (two [2%] patients), diarrhoea (two [2%] patients), and nausea and vomiting (one [1%] patient). According to imaging examination, 74 (75%) patients showed bilateral pneumonia, 14 (14%) patients showed multiple mottling and ground-glass opacity, and one (1%) patient had pneumothorax. 17 (17%) patients developed acute respiratory distress syndrome and, among them, 11 (11%) patients worsened in a short period of time and died of multiple organ failure.\nINTERPRETATION: The 2019-nCoV infection was of clustering onset, is more likely to affect older males with comorbidities, and can result in severe and even fatal respiratory diseases such as acute respiratory distress syndrome. In general, characteristics of patients who died were in line with the MuLBSTA score, an early warning model for predicting mortality in viral pneumonia. Further investigation is needed to explore the applicability of the MuLBSTA score in predicting the risk of mortality in 2019-nCoV infection.\nFUNDING: National Key R&D Program of China.'

SUMMARIZED:

'2019-nCoV pneumonia was detected in all patients by real-time RT-PCR . 49 (49%) had a history of exposure to the Huanan seafood market . the average age of the patients was 55.5 years (SD 131), including 67 men and 32 women . INTERPRETATION: 2018-ncoV infection is more likely to affect older males .'

(2) ORIGINAL:

'Importance: The outbreak of coronavirus disease 2019 (COVID-19) in Wuhan, China, is serious and has the potential to become an epidemic worldwide. Several studies have described typical clinical manifestations including fever, cough, diarrhea, and fatigue. However, to our knowledge, it has not been reported that patients with COVID-19 had any neurologic manifestations.\nObjective: To study the neurologic manifestations of patients with COVID-19.\nDesign, Setting, and Participants: This is a retrospective, observational case series. Data were collected from January 16, 2020, to February 19, 2020, at 3 designated special care centers for COVID-19 (Main District, West Branch, and Tumor Center) of the Union Hospital of Huazhong University of Science and Technology in Wuhan, China. The study included 214 consecutive hospitalized patients with laboratory-confirmed diagnosis of severe acute respiratory syndrome coronavirus 2 infection.\nMain Outcomes and Measures: Clinical data were extracted from electronic medical records, and data of all neurologic symptoms were checked by 2 trained neurologists. Neurologic manifestations fell into 3 categories: central nervous system manifestations (dizziness, headache, impaired consciousness, acute cerebrovascular disease, ataxia, and seizure), peripheral nervous system manifestations (taste impairment, smell impairment, vision impairment, and nerve pain), and skeletal muscular injury manifestations.\nResults: Of 214 patients (mean [SD] age, 52.7 [15.5] years; 87 men [40.7%]) with COVID-19, 126 patients (58.9%) had nonsevere infection and 88 patients (41.1%) had severe infection according to their respiratory status. Overall, 78 patients (36.4%) had neurologic manifestations. Compared with patients with nonsevere infection, patients with severe infection were older, had more underlying disorders, especially hypertension, and showed fewer typical symptoms of COVID-19, such as fever and cough. Patients with more severe infection had neurologic manifestations, such as acute cerebrovascular diseases (5 [5.7%] vs 1 [0.8%]), impaired consciousness (13 [14.8%] vs 3 [2.4%]), and skeletal muscle injury (17 [19.3%] vs 6 [4.8%]).\nConclusions and Relevance: Patients with COVID-19 commonly have neurologic manifestations. During the epidemic period of COVID-19, when seeing patients with neurologic manifestations, clinicians should suspect severe acute respiratory syndrome coronavirus 2 infection as a differential diagnosis to avoid delayed diagnosis or misdiagnosis and lose the chance to treat and prevent further transmission.'

SUMMARIZED:

'the outbreak of coronavirus disease 2019 (COVID-19) in Wuhan, china, is serious and has the potential to become an epidemic worldwide . the study included 214 consecutive hospitalized patients with laboratory-confirmed diagnosis of severe acute respiratory syndrome coroniavirus 2 infection . 78 patients (36.4%) had neurologic manifestations, including fever, cough, diarrhea, and fatigue .'

(3) ORIGINAL:

'The clinical features and immune responses of asymptomatic individuals infected with severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) have not been well described. We studied 37 asymptomatic individuals in the Wanzhou District who were diagnosed with RT–PCR-confirmed SARS-CoV-2 infections but without any relevant clinical symptoms in the preceding 14 d and during hospitalization. Asymptomatic individuals were admitted to the government-designated Wanzhou People’s Hospital for centralized isolation in accordance with policy1. The median duration of viral shedding in the asymptomatic group was 19 d (interquartile range (IQR), 15–26 d). The asymptomatic group had a significantly longer duration of viral shedding than the symptomatic group (log-rank P\u2009=\u20090.028). The virus-specific IgG levels in the asymptomatic group (median S/CO, 3.4; IQR, 1.6–10.7) were significantly lower (P\u2009=\u20090.005) relative to the symptomatic group (median S/CO, 20.5; IQR, 5.8–38.2) in the acute phase. Of asymptomatic individuals, 93.3% (28/30) and 81.1% (30/37) had reduction in IgG and neutralizing antibody levels, respectively, during the early convalescent phase, as compared to 96.8% (30/31) and 62.2% (23/37) of symptomatic patients. Forty percent of asymptomatic individuals became seronegative and 12.9% of the symptomatic group became negative for IgG in the early convalescent phase. In addition, asymptomatic individuals exhibited lower levels of 18 pro- and anti-inflammatory cytokines. These data suggest that asymptomatic individuals had a weaker immune response to SARS-CoV-2 infection. The reduction in IgG and neutralizing antibody levels in the early convalescent phase might have implications for immunity strategy and serological surveys.'

SUMMARIZED:

'37 asymptomatic individuals in the wanzhou district were diagnosed with RT–PCR-confirmed SARS-CoV-2 infections . they were admitted to the government-designated Wanzhou People’s Hospital for centralized isolation in accordance with policy1 . median duration of viral shedding was 19 d (interquartile range (IQR), 15–26 d)'

How we built it

We used state-of-the-art machine learning techniques including transformers to read in thousands of the most highly cited COVID-19 related publications since 2019, read their abstracts and generate summaries of the dense scientific text in easy-to-understand vernacular. This was done using Hugging Face summarizer pipeline, a pre-trained transformer-based model. We built out a MongoDB backend to store the COVID-19 publication information and the summaries we generated. We used create-react-app and semantic ui to build out the front end and connected the two using express-graphql.

Challenges we ran into

The first ML model we tried to use could not take in the shape of the data we needed and therefore we could not train it. We spent a lot of time trying to get Pegasus to work, but were unable to. The model was also massive so we were using GPUs, and our current method with Hugging Face works on a regular computer. Other issues came with formatting graphQL queries and mutations and figuring out how to convert the data from the ML model to fit the shape of our MongoDB schema.

Accomplishments that we're proud of

Finishing on time! And really all the thought and time that we managed to put into our project :)

What we learned

Lots about pytorch, tensorflow, transformers, graphQL, react useEffect and useState hooks!

What's next for Publications for the Public (P4P)

Future directions for P4P include providing a new set of keyword filters such as what data was used i.e. scRNAseq, RNAseq, etc. and filter by what techniques were applied to analyze the data (immunohistochemistry, machine learning etc.)

Users could also be able to input text based on information they heard like "you can only quarantine for 4 days" and have relevant articles come up to combat that misinformation. Potentially the webapp could mine twitter for misinformation tweets.

References

Resources, Dimensions (2020): Dimensions COVID-19 publications, datasets and clinical trials. Dimensions. Dataset. https://doi.org/10.6084/m9.figshare.11961063.v38

Built With

Share this project:

Updates