Mumemto and Shredtools HPRC indexes

Mumemto (GitHub repo) is an efficient multi-MUM finder for large pangenomes. Multi-MUMs have a variety of uses, including forming a pangenome coordinate system, visualizing synteny, and serving as a basis for multiple genome alignment.

Shredtools (GitHub repo) is a companion tool that operates over multi-MUMs to navigate the pangenome coordinate system. Shredtools can extract homologous regions across a pangenome using multi-MUMs as a framework for the underlying multiple alignment.

We provide pre-computed multi-MUMs for a collection of 476 high-quality human genome assemblies (available from https://github.com/lh3/OpenHGL). The formats are as follows:

1. *.bumbl - native binary format for Mumemto. Fast loading into memory for downstream applications (Mumemto python API available), and can stream to human-readable *.mums with mumemto convert.

   Note: *.bumbl files are indexable with a *.bumbl.bi file (introduced with Shredtools). This enables retrieving the subset of multi-MUMs contained in a requested region from any assembly in the collection.

2. *.mums.gz - gzip of human readable *.mums file, sorted by occurrence in the first genome (CHM13).
3. *.lengths - a manifest of contig lengths for each genome, used in multiple downstream Mumemto scripts, and dictates the order of genomes / contigs in the MUM file.
4. hprcv2.athresh - threshold file necessary to merge in new genomes to the pangenome collection.

The following files contain strict multi-MUMs, exact matches that appear in every genome, exactly once per assembly. This is useful for building alignments, graphs, and identifying variation.

Corresponding S3 URLs can be obtained by replacing the https://genome-idx.s3.amazonaws.com/ prefix with s3://genome-idx/.

Note: hprcv2.lengths can be used with any of the mums/bumbl files, since it contains a manifest of assembly and contig lengths in the collection. hprcv2.athresh can also be used to merge in new assemblies, if mumemto was run between the CHM13 and the new assembly.

In the following files, we merge consecutive multi-MUMs separated by an SNV. Lines in these files may contain single base mismatches, but no indels. This is useful for identifying pangenome coordinates and syntenic blocks with fewer (but longer) match blocks. These files were generated post-filtering out non-collinear matches.

Shredtools HPRC indexes

Shredtools uses multi-MUMs to slice and extract homologous regions. See the interactive webapp version at https://vikshiv.github.io/shredtools/, which uses the following files:

Note: FASTA sequences are not hosted alongside these index files. Shredtools retrieves genome sequences directly from the HPRC assembly collection (s3://human-pangenomics/). The interactive webapp version can also fetch queried sequences from the HPRC S3 bucket.

Mumemto and Shredtools are the work of Vikram Shivakumar and Ben Langmead.