SICaRiO

SICaRiO enhances detection of true insertions and deletions (indels) in next-generation sequencing (NGS) data to reduce false positives in variant calling.

Key Features:

• Gradient Boosting Classifier: SICaRiO employs a gradient boosting machine learning classifier to classify indel calls.
• Training on Gold-Standard Data: The classifier is trained on the Genome in a Bottle (GIAB) consortium gold-standard dataset.
• Independence from Pipeline-Specific Metrics: The method does not require pipeline-specific information such as read depth and instead uses genomic features derivable from public resources.
• Improved Variant Calling Performance: SICaRiO filters false positive indel calls to improve the performance of variant calling pipelines.
• Genomic Context Insight: The tool analyzes prior genomic contexts that contribute to erroneous indel calls.
• Prediction Difficulty Analysis: SICaRiO compares prediction difficulty across three categories of indels and across sequencing pipelines.
• Feature Ranking for False Positives: The method ranks genomic features by their predictivity for false positive indel calls.

Scientific Applications:

• Personalized Healthcare: Improves accuracy of indel detection for clinical and personalized genomic analyses.
• Disease Genomics: Aids identification of disease-associated indels for research into genetic causes of disease.
• Population Genetics: Enables more accurate studies of genetic diversity and evolutionary patterns by reducing indel call errors.

Methodology:

SICaRiO uses a gradient boosting machine learning classifier trained on GIAB gold-standard data, leverages genomic features derivable from public resources rather than pipeline-specific metrics, compares prediction difficulty across three indel categories and sequencing pipelines, and ranks genomic features by their predictivity for false positives.