Deep-GenMut: Automated genetic mutation classification in oncology: A deep learning comparative study.

Area of Science:

• Computational biology
• Bioinformatics
• Genomics

Background:

• Manual classification of genetic mutations by pathologists is time-consuming and can impact early cancer detection.
• Next-generation sequencing technologies enable automated mutation analysis, improving clinical interpretation precision.

Purpose of the Study:

• To evaluate deep learning models for automated classification of genetic mutations using biomedical texts.
• To compare the performance of BioBERT against other models like BERT, LSTM, and BiLSTM for this task.

Main Methods:

• Utilized four deep learning classification models: BioBERT, BERT, LSTM, and BiLSTM.
• Trained models on a dataset of biomedical texts containing genetic mutations, sourced from Memorial Sloan Kettering Cancer Center.
• Addressed challenges including enormous text length, data bias, and data repetition.

Main Results:

• BioBERT demonstrated superior performance, achieving an F1 score of 0.87 and a Matthews Correlation Coefficient (MCC) of 0.850.
• This represents a significant improvement over the BERT model, which achieved an F1 score of 0.70.
• The models were evaluated using standard metrics on a challenging Kaggle dataset.

Conclusions:

• BioBERT is a highly effective model for automated genetic mutation classification from biomedical texts.
• The developed computational approach enhances the accuracy and efficiency of mutation analysis for early cancer detection.
• Further research can build upon BioBERT's success to refine automated clinical interpretation tools.