ralphi: a deep reinforcement learning framework for haplotype assembly.

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• Haplotype assembly is crucial for understanding how allele combinations affect phenotype.
• Reconstructing haplotypes from individual diploid genomes is essential for genetic studies.
• Existing methods face challenges in accurately partitioning DNA reads into their respective haplotype sets.

Purpose of the Study:

• To introduce ralphi, a novel deep reinforcement learning framework for read-based haplotype assembly.
• To improve the accuracy and efficiency of reconstructing haplotypes from diploid genomes.
• To leverage deep learning and reinforcement learning for precise read fragment partitioning.

Main Methods:

• Developed ralphi, a deep reinforcement learning framework integrating deep learning and reinforcement learning.
• Utilized the maximum fragment cut formulation on fragment graphs for reinforcement learning reward objectives.
• Trained ralphi on diverse fragment graph topologies from the 1000 Genomes Project.

Main Results:

• ralphi demonstrates consistently lower error rates compared to state-of-the-art methods.
• Achieved comparable or longer haplotype block lengths in human genome benchmarks.
• Showed robust performance across short and long ONT reads at varying coverage levels.

Conclusions:

• ralphi offers a significant advancement in read-based haplotype assembly.
• The deep reinforcement learning approach enhances accuracy in reconstructing diploid haplotypes.
• ralphi provides a valuable tool for genomic research, improving variant impact analysis.