Training deep learning models on personalized genomic sequences improves variant effect prediction.

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• Sequence-to-function models are crucial for understanding genetic variation's molecular effects.
• Current models face challenges with accuracy in variant effect prediction.
• Interpreting trait-associated genetic variation remains a significant challenge.

Purpose of the Study:

• To enhance the performance of sequence-to-function models for variant effect prediction.
• To investigate the impact of training data on model accuracy.
• To assess the generalizability of learned variant effect representations.

Main Methods:

• Training sequence-to-function models using functional genomic data.
• Integrating matched personal genomes into the training process.
• Fine-tuning models on diverse cellular contexts and experimental readouts.

Main Results:

• Models trained on functional genomic data with matched personal genomes showed improved variant effect prediction.
• Learned variant effect representations were robust and retained when fine-tuning.
• Enhanced model performance has implications for understanding genetic variation and traits.

Conclusions:

• Integrating personal genomes with functional genomic data is key to improving sequence-to-function models.
• The developed models offer a more accurate approach to variant effect prediction.
• This work facilitates better interpretation of genetic variation influencing human traits.