A differential diagnosis between uterine leiomyoma and leiomyosarcoma using transcriptome analysis

Area of Science:

• Oncology
• Genomics
• Bioinformatics

Background:

• Distinguishing uterine leiomyoma from leiomyosarcoma is crucial for appropriate treatment.
• Transcriptome data offers a potential avenue for accurate differential diagnosis.

Purpose of the Study:

• To develop and validate a transcriptome-based classifier for differentiating uterine leiomyosarcoma from leiomyoma.
• To assess the accuracy of machine learning models using selected gene expression profiles.

Main Methods:

• Utilized publicly available transcriptome data for training and validation sets.
• Developed a gene selection method to identify genes with higher variance in leiomyosarcoma.
• Trained and evaluated deep feedforward neural network (DNN), SVM, RF, and GB models.
• Validated the classifier on 35 newly collected clinical samples.

Main Results:

• Identified 17 genes with variable expression in leiomyosarcoma compared to normal uterine tissues, primarily associated with DNA replication.
• The DNN classifier demonstrated high accuracy, with sensitivity ranging from 0.77 to 0.88 and specificity of 1.0 in clinical samples.
• The classifier effectively discriminated between leiomyosarcoma and leiomyoma in independent test sets.

Conclusions:

• A transcriptome-based classifier accurately distinguishes uterine leiomyosarcoma from leiomyoma.
• This approach can assist in pre-surgical diagnosis via biopsy, guiding treatment decisions.
• Accurate identification can prevent unnecessary surgeries and reduce the risk of tumor spread.