Performance evaluation of differential splicing analysis methods and splicing analytics platform construction

Area of Science:

• Genetics
• Bioinformatics
• Molecular Biology

Background:

• Human genetic variants of uncertain significance can cause abnormal splicing.
• Existing splicing variant prediction methods lack comprehensive evaluation on independent benchmarks.

Purpose of the Study:

• To evaluate the performance of existing splicing variant prediction methods using a manually curated, independent dataset.
• To introduce SPCards, a novel platform for curating and predicting splicing variants.

Main Methods:

• Manually curated a dataset of ~50,000 splicing variants from over 8,000 studies.
• Evaluated performance metrics (sensitivity, specificity) of various splicing prediction tools.
• Developed the SPCards platform for variant curation and splicing score prediction.

Main Results:

• Prediction methods demonstrated varied performance in donor and acceptor regions.
• A majority of methods (66.67%) showed higher specificity than sensitivity, indicating a need for adjusted cut-off values.
• High correlation and consistent prediction ratios supported the integration of splicing prediction methods.
• SPCards platform provides variant and gene-level annotations for high-throughput genetic identification.

Conclusions:

• Integration of splicing prediction methods is feasible for identifying splicing variants.
• SPCards facilitates high-throughput genetic identification of splicing variants, including those in non-canonical regions.
• Further refinement of prediction models and cut-off values is suggested for improved accuracy.