State-of-the-art fusion-finder algorithms sensitivity and specificity.

Area of Science:

• Genomics
• Bioinformatics
• Cancer Research

Background:

• Chromosomal translocations can lead to gene fusions, driving cancer development.
• RNA sequencing (RNA-seq) is a powerful tool for identifying these rearrangements and resultant chimeric proteins.
• Existing chimera detection tools lack comprehensive evaluation of their specificity and sensitivity.

Purpose of the Study:

• To comparatively evaluate the performance of eight RNA-seq-based chimera detection tools.
• To assess the reliability of synthetic versus real datasets for chimera detection tool benchmarking.
• To identify limitations in current fusion-finding algorithms and suggest areas for improvement.

Main Methods:

• Eight fusion-detection tools were tested: FusionHunter, FusionMap, FusionFinder, MapSplice, deFuse, Bellerophontes, ChimeraScan, and TopHat-fusion.
• Both synthetic and real RNA-seq datasets containing known chimeras were utilized for evaluation.
• Novel filtering strategies were developed and applied to reduce false positives from the ChimeraScan tool.

Main Results:

• Comparative analysis using only synthetic data yielded misleading results compared to real datasets.
• Most evaluated tools exhibited a high rate of false positive chimera predictions.
• The ChimeraScan tool, while sensitive, produced numerous false positives that were significantly reduced by applying custom filters for junction-spanning reads and intronic regions.

Conclusions:

• Synthetic datasets may not accurately reflect the complexities of real-world RNA-seq experiments for chimera detection.
• Current fusion detection tools possess limitations in both sensitivity and specificity.
• Further advancements in fusion-finder algorithms are necessary to improve accurate detection of cancer-associated gene fusions.