Movi indexes
Movi (GitHub repo) is an efficient and scalable approach for indexing and querying pangenomes. It uses the move structure of Nishimoto and Tabei as its core, cache-efficient index structure.
The indexes below are separator-aware, reverse-complement-aware pangenome indexes for Movi 2. They are built in Movi’s regular-thresholds mode (mode 6) and include threshold information, so they support the full range of Movi 2 queries: pseudo-matching lengths (--pml), Ziv–Lempel matching lengths (--zml), --count, k-mer presence/counts (--kmer), maximal exact matches (--mem), and — using the shipped null model — read --classify and --filter. A % separator is placed after every record and between each record’s forward and reverse-complement copy, so matches never spuriously extend across a record (contig) boundary.
Each index was built on JHU’s Rockfish cluster. The two larger indexes (HPRC Release 2 and OpenHGL) are the first Movi indexes whose reference length exceeds 240 bp; they use a wide (6-byte) threshold format produced by an updated build path. The published indexes are queryable by stock Movi 2 — the wide-threshold change is build-time only and does not alter the on-disk index format.
Every index ships a signed test-witness.json recording the results of a corner-case query-mode test suite (separator non-extension, reverse-complement symmetry, illegal/non-ACGT character handling, positive-vs-null separation, and cross-mode consistency) run against that exact index.movi; the witness pins the index’s SHA-256 and the Movi commit, and carries a self-verifying self_sha256. All three indexes passed with zero failures.
The following table lists each complete .tar.gz archive, then the individual component files contained in that archive.
| Index / file | Size | HTTPS |
|---|---|---|
HPRC Release 1 (Year 1), 95 human haplotypes (complete .tar.gz) |
18.9 GB | .tar.gz |
| Link above is recommended; links below are its individual components | ||
index.movi |
37.6 GB | .movi |
ftab.2.bin |
528 B | .bin |
ftab.4.bin |
8.2 KB | .bin |
ftab.6.bin |
131 KB | .bin |
ftab.8.bin |
2.1 MB | .bin |
ftab.10.bin |
33.6 MB | .bin |
ftab.12.bin |
537 MB | .bin |
movi.pml.nulldb |
1.2 MB | .nulldb |
movi.zml.nulldb |
1.2 MB | .nulldb |
null_reads.fasta |
161 KB | .fasta |
test-witness.json |
10 KB | .json |
movi-index.yaml |
1.1 KB | .yaml |
hprc-yr1.seqdict.json |
4.9 MB | .json |
HPRC Release 2 (Year 2), 472 human haplotypes (complete .tar.gz) |
26.1 GB | .tar.gz |
| Link above is recommended; links below are its individual components | ||
index.movi |
48.6 GB | .movi |
ftab.2.bin |
528 B | .bin |
ftab.4.bin |
8.2 KB | .bin |
ftab.6.bin |
131 KB | .bin |
ftab.8.bin |
2.1 MB | .bin |
ftab.10.bin |
33.6 MB | .bin |
ftab.12.bin |
537 MB | .bin |
movi.pml.nulldb |
1.2 MB | .nulldb |
movi.zml.nulldb |
1.2 MB | .nulldb |
null_reads.fasta |
161 KB | .fasta |
test-witness.json |
10 KB | .json |
movi-index.yaml |
1.1 KB | .yaml |
hprc-yr2.seqdict.json |
5.0 MB | .json |
OpenHGL, 579 human haplotypes (complete .tar.gz) |
28.1 GB | .tar.gz |
| Link above is recommended; links below are its individual components | ||
index.movi |
51.6 GB | .movi |
ftab.2.bin |
528 B | .bin |
ftab.4.bin |
8.2 KB | .bin |
ftab.6.bin |
131 KB | .bin |
ftab.8.bin |
2.1 MB | .bin |
ftab.10.bin |
33.6 MB | .bin |
ftab.12.bin |
537 MB | .bin |
movi.pml.nulldb |
1.2 MB | .nulldb |
movi.zml.nulldb |
1.2 MB | .nulldb |
null_reads.fasta |
161 KB | .fasta |
test-witness.json |
10 KB | .json |
movi-index.yaml |
1.1 KB | .yaml |
openhgl.seqdict.json |
6.4 MB | .json |
The .seqdict.json file maps each index position back to its source record (sample, contig, and orientation). The movi-index.yaml sidecar records the index header (mode, reference length, run count) and build provenance.
Corresponding S3 URLs can be obtained by replacing the https://genome-idx.s3.amazonaws.com/ prefix with s3://genome-idx/.
Movi is the work of Mohsen Zakeri, Nathaniel Brown, Travis Gagie and Ben Langmead.
