liftover

Liftover is the process of converting genomic coordinates or annotations from one reference genome assembly to another. It is used when a genome assembly has been updated or replaced with a more accurate version, so that coordinates for genes, variants, and other features remain comparable across studies.

The method relies on chain files or equivalent alignment mappings that describe how regions in the source assembly align to the target assembly. Tools take input coordinates (for example, BED intervals or VCF variant positions) and output the corresponding coordinates in the target assembly. Results can be successful, partially successful, or fail entirely for regions that cannot be unambiguously mapped. In some cases, information may be ambiguous or duplicated, leading to multiple possible mappings.

Common tools and formats include UCSC liftOver, CrossMap, and Picard’s LiftOverVcf. These tools require a chain file that encodes pairwise alignments between assemblies and typically support various input formats such as BED, VCF, or GFF. The process is designed to preserve the biological meaning of coordinates as much as possible, while recognizing that exact one-to-one mapping is not guaranteed for all regions.

Limitations and caveats include regions that have undergone rearrangements, insertions, or deletions between assemblies, which may not map uniquely or at all. Chain file quality and coverage directly affect accuracy, and liftover can introduce coordinate errors if downstream analyses assume exact positions. Liftover is often preferred to avoid re-aligning large datasets, but in some cases re-alignment to the new assembly may be necessary for optimal results.

Applications include updating variant catalogs to a newer assembly, harmonizing annotations across studies, and facilitating meta-analyses that combine data generated on different reference genomes.