LBFextract

LBFextract streamlines feature extraction from circulating cell-free DNA (cfDNA) data, integrating preprocessing and postprocessing steps to enable biomarker discovery by inferring transcription factor activity.

Key Features:

• Coverage-based feature extraction: Extracts features from cfDNA data using coverage-based approaches at genomic regions of interest.
• Fragment length-based feature extraction: Extracts features from cfDNA data using fragment length-based approaches.
• Entropy-based TF activity inference: A novel method to infer transcription factor activity from fragmentomics data using entropy-based calculations.
• Nucleosome dyad signal amplification: A technique to amplify signals from nucleosome dyads in cfDNA fragment data.
• Condition-specific differential TF activity: Implements a method to extract condition-specific differentially active transcription factors based on extracted features for biomarker discovery.
• Alignment fragment tag integration: Facilitates integration of preprocessing and postprocessing steps through alignment fragment tags.
• Hook mechanism: Enables extensible integration of preprocessing and postprocessing steps via a hook mechanism.

Scientific Applications:

• Transcription factor activity inference: Infers transcription factor dynamics and activity from liquid biopsy cfDNA fragmentomics data.
• Biomarker discovery: Identifies condition-specific differentially active transcription factors as potential biomarkers from cfDNA features.
• Prostate cancer subtype classification: Demonstrated for subtype classification of advanced prostate cancer patients using coverage signals at transcription factor binding sites from cfDNA.
• Clinical group discrimination: Generates robust and interpretable features that discriminate between different clinical groups.
• Liquid biopsy study reproducibility: Enhances reproducibility and comparability of liquid biopsy studies through standardized feature extraction.

Methodology:

LBFextract processes cfDNA data through alignment fragment tags with a hook mechanism for preprocessing and postprocessing integration. It implements coverage-based and fragment length-based approaches, an entropy-based method for TF activity inference from fragmentomics data, a technique for nucleosome dyad signal amplification, and a method to extract condition-specific differentially active TFs.