No abstract

Medicines based on messenger RNA (mRNA) hold immense potential, as evidenced by their rapid deployment as COVID-19 vaccines. However, worldwide distribution of mRNA molecules has been limited by their thermostability, which is fundamentally limited by the intrinsic instability of RNA molecules to a chemical degradation reaction called in-line hydrolysis. Predicting the degradation of an RNA molecule is a key task in designing more stable RNA-based therapeutics. Here, we describe a crowdsourced machine learning competition (‘Stanford OpenVaccine’) on Kaggle, involving single-nucleotide resolution measurements on 6,043 diverse 102–130-nucleotide RNA constructs that were themselves solicited through crowdsourcing on the RNA design platform Eterna. The entire experiment was completed in less than 6 months, and 41% of nucleotide-level predictions from the winning model were within experimental error of the ground truth measurement. Furthermore, these models generalized to blindly predicting orthogonal degradation data on much longer mRNA molecules (504–1,588 nucleotides) with improved accuracy compared with previously published models. These results indicate that such models can represent in-line hydrolysis with excellent accuracy, supporting their use for designing stabilized messenger RNAs. The integration of two crowdsourcing platforms, one for dataset creation and another for machine learning, may be fruitful for other urgent problems that demand scientific discovery on rapid timescales.

More and more biomedical documents are digitally written and stored. To make the most of the rich resources, it is crucial to precisely locate the information pertinent to users' interests. One of the obstacles in finding information in natural language text is negations, which deny or reverse the meaning of a sentence. This is especially problematic in the biomedical domain since scientific findings and clinical records often contain negated expressions to state negative effects or the absence of symptoms. Ignoring such negated expressions result in more irrelevant information and may even lead to false conclusions. Therefore, identifying negative words and their scopes are important sub-tasks in biomedical information processing. This paper reports on our ongoing work on a hybrid approach to negation identification combining statistical and heuristic approaches. Our approach is evaluated on three types of biomedical documents in comparison with an existing machine learning approach. In addition, the empirical results are manually analyzed to better understand the nature of the problems.

More and more biomedical documents are digitally written and stored. To make the most of the rich resources, it is crucial to precisely locate the information pertinent to user's interests. An obstacle in finding information in natural language text is negations, which deny or reverse the meaning of a sentence. This is especially problematic in the biomedical domain since scientific findings and clinical records often contain negated expressions to state negative effects or the absence of symptoms. This paper reports on our work on a hybrid approach to negation identification combining statistical and heuristic approaches and describes an implementation of the approach, named NegFinder, as a Web service.

No abstract