Scholarlink

<p>Variants of coronavirus disease 2019 (COVID-19) may exhibit differing levels of virulence as the virus shifts from pandemic to endemic status. To prepare for anticipated challenges to hospital capacity and protect vulnerable populations, it is necessary to predict whether a newly emerged COVID-19 variant carries an increased risk of causing severe disease. This work intends to devise a Fuzzy-based maxout–Long Short-Term Memory (FMLSTM) based variant impact predictions utilizing the Severe Acute Respiratory Syndrome Corona Virus 2 (SARS-CoV2) database. Primarily, SARS-CoV2 data is acquired as input, and translation of Non-Structural Proteins (NSPs) is done. Then, the protein sequence is observed, and from this, the sequence-based features like the length of the sequence, Amino Acid count (AA), aromaticity, secondary structure fraction, and entropy are mined. Finally, the variant impact score is predicted based on a hybrid Deep Learning (DL) model named FMLSTM. Moreover, FMLSTM is developed by incorporating the Fuzzy concept, Deep Maxout Network (DMN), and Long Short-Term Memory (LSTM). The proposed FMLSTM outperformed various traditional methods and achieved effective performance with the minimum Mean Squared Error (MSE) of 8.3%, Mean Absolute Error (MAE) of 16.1%, and Root Mean Square Error (RMSE) of 28.9% with 90% of learning data.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Fuzzy-Based Maxout–LSTM Based Variant Impact Predictions Using SARS-CoV2 Database

  • 期刊论文

  • 发表日期：

    2026年4月29日

查看全文
  • Divya,
  • Alankrita Aggarwal,
  • Balajee Maram

摘要

Variants of coronavirus disease 2019 (COVID-19) may exhibit differing levels of virulence as the virus shifts from pandemic to endemic status. To prepare for anticipated challenges to hospital capacity and protect vulnerable populations, it is necessary to predict whether a newly emerged COVID-19 variant carries an increased risk of causing severe disease. This work intends to devise a Fuzzy-based maxout–Long Short-Term Memory (FMLSTM) based variant impact predictions utilizing the Severe Acute Respiratory Syndrome Corona Virus 2 (SARS-CoV2) database. Primarily, SARS-CoV2 data is acquired as input, and translation of Non-Structural Proteins (NSPs) is done. Then, the protein sequence is observed, and from this, the sequence-based features like the length of the sequence, Amino Acid count (AA), aromaticity, secondary structure fraction, and entropy are mined. Finally, the variant impact score is predicted based on a hybrid Deep Learning (DL) model named FMLSTM. Moreover, FMLSTM is developed by incorporating the Fuzzy concept, Deep Maxout Network (DMN), and Long Short-Term Memory (LSTM). The proposed FMLSTM outperformed various traditional methods and achieved effective performance with the minimum Mean Squared Error (MSE) of 8.3%, Mean Absolute Error (MAE) of 16.1%, and Root Mean Square Error (RMSE) of 28.9% with 90% of learning data.