A method and system for virus detection based on tumor RNA sequencing data

By using multimodal deep learning methods to extract features and predict tumor RNA sequencing data, this approach solves the problems of low efficiency and low accuracy in virus identification in existing technologies, achieving high-precision virus monitoring and identification, and is applicable to tumor sequencing data analysis in medical and research fields.

CN117746985BActive Publication Date: 2026-06-30XIAMEN UNIV
View PDF 2 Cites 0 Cited by

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
XIAMEN UNIV
Filing Date
2023-12-26
Publication Date
2026-06-30

AI Technical Summary

Technical Problem

Existing technologies are inefficient and inaccurate in identifying viruses in tumor sequencing data, making it difficult to discover new virus species, especially when dealing with large-scale and complex tumor omics data.

Method used

A multimodal deep learning approach is employed to preprocess tumor RNA sequencing data, extract features based on sequence information and codons, construct a prediction model based on sequence information and codons, and combine attention mechanisms and multi-head attention mechanisms for virus probability prediction and assembly.

Benefits of technology

It improves the accuracy and robustness of virus monitoring, can adaptively process sequencing data of different sources and lengths, effectively identify new potential viruses, expands the application field, and meets medical and research needs.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN117746985B_ABST
    Figure CN117746985B_ABST
Patent Text Reader

Abstract

A method and system for virus detection in tumor RNA sequencing data are disclosed. The method includes preprocessing the raw tumor RNA sequencing data; inputting the processed tumor RNA sequencing data into sequence-information-based channels and codon-based channels for feature extraction to generate a feature matrix; constructing a sequence information prediction model and a codon information prediction model, and inputting the feature matrices generated from the sequence information-based channels and codon-based channels into the sequence information prediction model and codon information prediction model, respectively, for training and optimization; predicting the virus probability of each sequencing read to obtain a model score; and selecting viral sequencing reads based on the model scores to assemble viral contigs. This invention improves the accuracy and robustness of virus monitoring by introducing a multimodal deep learning method, and can adaptively process sequencing data from different sources and of different lengths, thereby better meeting the needs of medical and research fields for virus identification in tumor sequencing data.
Need to check novelty before this filing date? Find Prior Art