Extracting and evaluating features from rna virus sequences to predict host species susceptibility using deep learning

Identifying and monitoring hosts of zoonotic RNA viruses, that is, RNA viruses which can be transmitted from one species to another, including the recent SARS-CoV-2 causing the COVID-19 pandemic, is paramount to control their spread. However, efforts to control such spread may be affected if there are unmonitored or unknown hosts. To help identify potential hosts that may harbour such zoonotic viruses, we propose a pipeline that extracts features from sequences of RNA viruses, then uses the extracted features with deep learning to predict host species susceptibility. In addition to using sequence-related features, our method also extracts and uses features derived from the RNA secondary structures that can be formed by the viral sequences, since RNA secondary structures are known to take part in virus-host interaction. We evaluated the performance of our method and the different extracted features with a dataset containing RNA virus sequences and the host they infect, regardless of the viral species, from the NCBI Virus database. Using 10-fold cross validation, we found that a combination of the extracted features yielded the highest overall prediction accuracy of 86.89%. © 2021 Owner/Author.

Sequence similarity analysis between SARS-COV-2 and other viruses based on variant logic framework and topological entropy

Severe Acute Respiratory Syndrome Coronavirus 2 (SARS-CoV-2) popular in just a few months the world, the present study found that the virus belongs to the β-coronavirus family. We study the sequence similarity, which can be coronavirus vaccine development and analysis provides a scientific help, including SARS-CoV-2 high similarity with Bat-CoV, SARS, etc. But the low accuracy of long time-consuming problems sequences analysis method. In this paper, the topological entropy of different combination dimensions of each sequence was calculated based on the Variant Logic Framework. the sequences of SARS CoV-2 and other categories of viruses were taken as input data. The sequence similarity matrix of mutual information among different sequences was obtained by calculating Euclidean distance. Finally, using a visualization diagram, generate the phylogenetic tree. The experimental results show that topology entropy is fast and effective for virus sequence processing and similarity analysis, which also provides a new idea for virus sequence research and traceability. © 2021 ACM.

Computational Prediction of N6-methyladenosine (m6A) RNA Methylation in SARS-CoV-2 Viral Transcripts

SARS-CoV-2 caused atypical pneumonia (COVID-19) is an ongoing pandemic that seriously threat the global public health. Many people die from this disease with severe symptoms. The most prevalent m6A RNA modification may be involved in by assisting the virus escaping from the host cell immune system attack. We provided here the first computational prediction study of RNA methylation sites in SARS-CoV-2. Based on virus sequence information, we predict the potential virus m6A sites and hope to make anyhow contributions to this unprecedented situation. As a result, we found 27 most frequent m6A sequences (41 bp) in SARS-CoV-2, and two of them are quite near to the spike protein stop codon position. © 2021 ACM.