Large-Scale Biomedical Relation Extraction Across Diverse Relation Types: Model Development and Usability Study on COVID-19

Integrating graph convolutional networks to enhance prompt learning for biomedical relation extraction

BACKGROUND AND OBJECTIVE

Biomedical relation extraction aims to reveal the relation between entities in medical texts. Currently, the relation extraction models that have attracted much attention are mainly to fine-tune the pre-trained language models (PLMs) or add template prompt learning, which also limits the ability of the model to deal with grammatical dependencies. Graph convolutional networks (GCNs) can play an important role in processing syntactic dependencies in biomedical texts.

METHODS

In this work, we propose a biomedical relation extraction model that fuses GCNs enhanced prompt learning to handle limitations in syntactic dependencies and achieve good performance. Specifically, we propose a model that combines prompt learning with GCNs for relation extraction, by integrating the syntactic dependency information analyzed by GCNs into the prompt learning model, by predicting the correspondence with [MASK] tokens labels for relation extraction.

RESULTS

Our model achieved F1 scores of 85.57%, 80.15%, 95.10%, and 84.11% in the biomedical relation extraction datasets GAD, ChemProt, PGR, and DDI, respectively, all of which outperform some existing baseline models.

CONCLUSIONS

In this paper, we propose enhancing prompt learning through GCNs, integrating syntactic information into biomedical relation extraction tasks. Experimental results show that our proposed method achieves excellent performance in the biomedical relation extraction task.

Influence of Environmental Factors and Genome Diversity on Cumulative COVID-19 Cases in the Highland Region of China: Comparative Correlational Study

BACKGROUND

The novel coronavirus SARS-CoV-2 caused the global COVID-19 pandemic. Emerging reports support lower mortality and reduced case numbers in highland areas; however, comparative studies on the cumulative impact of environmental factors and viral genetic diversity on COVID-19 infection rates have not been performed to date.

OBJECTIVE

The aims of this study were to determine the difference in COVID-19 infection rates between high and low altitudes, and to explore whether the difference in the pandemic trend in the high-altitude region of China compared to that of the lowlands is influenced by environmental factors, population density, and biological mechanisms.

METHODS

We examined the correlation between population density and COVID-19 cases through linear regression. A zero-shot model was applied to identify possible factors correlated to COVID-19 infection. We further analyzed the correlation of meteorological and air quality factors with infection cases using the Spearman correlation coefficient. Mixed-effects multiple linear regression was applied to evaluate the associations between selected factors and COVID-19 cases adjusting for covariates. Lastly, the relationship between environmental factors and mutation frequency was evaluated using the same correlation techniques mentioned above.

RESULTS

Among the 24,826 confirmed COVID-19 cases reported from 40 cities in China from January 23, 2020, to July 7, 2022, 98.4% (n=24,430) were found in the lowlands. Population density was positively correlated with COVID-19 cases in all regions (ρ=0.641, P=.003). In high-altitude areas, the number of COVID-19 cases was negatively associated with temperature, sunlight hours, and UV index (P=.003, P=.001, and P=.009, respectively) and was positively associated with wind speed (ρ=0.388, P<.001), whereas no correlation was found between meteorological factors and COVID-19 cases in the lowlands. After controlling for covariates, the mixed-effects model also showed positive associations of fine particulate matter (PM2.5) and carbon monoxide (CO) with COVID-19 cases (P=.002 and P<.001, respectively). Sequence variant analysis showed lower genetic diversity among nucleotides for each SARS-CoV-2 genome (P<.001) and three open reading frames (P<.001) in high altitudes compared to 300 sequences analyzed from low altitudes. Moreover, the frequencies of 44 nonsynonymous mutations and 32 synonymous mutations were significantly different between the high- and low-altitude groups (P<.001, mutation frequency>0.1). Key nonsynonymous mutations showed positive correlations with altitude, wind speed, and air pressure and showed negative correlations with temperature, UV index, and sunlight hours.

CONCLUSIONS

By comparison with the lowlands, the number of confirmed COVID-19 cases was substantially lower in high-altitude regions of China, and the population density, temperature, sunlight hours, UV index, wind speed, PM2.5, and CO influenced the cumulative pandemic trend in the highlands. The identified influence of environmental factors on SARS-CoV-2 sequence variants adds knowledge of the impact of altitude on COVID-19 infection, offering novel suggestions for preventive intervention.