Integrating genetic variation with deep learning provides context for variants impacting transcription factor binding...

Area of Science:

• Genomics
• Molecular Biology
• Developmental Biology

Background:

• Understanding genetic variation's impact on transcription factor (TF) binding is crucial for modeling diseases.
• Current methods face challenges in accurately profiling allele-specific TF binding.

Purpose of the Study:

• To investigate how genetic variation affects TF binding during *Drosophila* embryogenesis.
• To develop and refine computational tools for detecting and interpreting allele-specific TF binding.

Main Methods:

• Utilized controlled F1 crosses with diverse genetic backgrounds for allele-specific TF binding profiling.
• Expanded the WASP tool to detect indels, enhancing the identification of allelically imbalanced TF binding peaks.
• Trained a convolutional neural network (Basenji) to predict TF binding from DNA sequence and model allelic imbalance.

Main Results:

• Identified 9%-18% of TF-bound regions affected by genetic variation, even for essential TFs.
• Increased detection of allelically imbalanced peaks by 30%-50% using the enhanced WASP tool.
• The Basenji model accurately predicted TF binding and allelic imbalance, offering mechanistic insights into variant effects.

Conclusions:

• Genetic variation plays a substantial role in regulating TF binding dynamics.
• Developed advanced computational approaches for precise identification and interpretation of genetic variants influencing TF binding.
• Uncovered novel TF relationships and tissue-specific binding mechanisms, including for CTCF.