LZGraphs

LZ76 and FlashBack compression graphs for immune receptor repertoire analysis

Documentation · Quick Start · API Reference · Report Bug

LZGraphs is a Python library that turns T-cell and B-cell receptor CDR3 sequences into probabilistic directed graphs. It ships two graph families on a shared C core:

LZGraph: built from Lempel-Ziv 76 compression. Supports V/J gene annotation, three encoding variants, and a lzg CLI.
FlashBackGraph: a Markovian DAG built from FlashBack tokenization (recursive run-peeling from both ends of the sentinel-wrapped sequence). Diversity, entropy, and path counting have closed-form forward-DP solutions; sequence simulation is still sampled.

Both classes share a common surface for scoring, simulation, diversity, graph algebra, posterior personalization, and binary serialization. See When to use which for a comparison.

Example LZGraph built from 3 CDR3 sequences
An LZGraph built from three CDR3s. @ and $ are start/end sentinels; subpattern nodes carry position suffixes.

Installation

pip install LZGraphs

Requires Python 3.9 or later. Wheels are published for Linux, macOS, and Windows (CPython 3.9–3.12). Release history: CHANGELOG.md.

A container image with the lzg CLI preinstalled is also published to GHCR:

docker run --rm -v "$PWD:/data" ghcr.io/mutejester/lzgraphs build /data/repertoire.tsv -o /data/repertoire.lzg

Input format

For programmatic use, all classes accept a plain list of CDR3 strings:

LZGraph(['CASSLEPSGGTDTQYF', 'CASSDTSGGTDTQYF', ...], variant='aap')

For files, LZGraph.from_file, FlashBackGraph.from_file, and the lzg CLI (lzg build / lzg flashback build) all read through the same format-detection pipeline and accept the same formats, so any of the three build the identical graph from the identical file:

Format	Layout	Example
Plain	one sequence per line	`CASSLEPSGGTDTQYF`
Seq + count	`sequence\tcount` (tab-separated)	`CASSLEPSGGTDTQYF\t42`
AIRR-compatible tabular	tab- or comma-separated, with header row	`junction_aa`, `v_call`, `j_call`, ...
FASTA	`>` header line, then sequence line(s)	`>seq1` / `CASSLEPSGGTDTQYF`
FASTQ	4-line records: header, sequence, `+`, quality	`@seq1` / `CASSLEPSGGTDTQYF` / `+` / `IIIIIIIIIIIIIII`

For AIRR-style tabular input: the sequence column is auto-detected from junction_aa / cdr3_amino_acid / cdr3_aa (variant aap), junction / cdr3_rearrangement (variant ndp), or any column named sequence/cdr3/seq. Gene calls come from v_call / j_call and must use IMGT-style notation (e.g. TRBV5-1*01); FlashBackGraph has no gene-annotation model, so a gene column present in the file is simply not read. Malformed sequence fields, and AIRR rows whose productive column is present and not truthy, are dropped and counted rather than built into the graph.

Compression is detected from file content, not the filename: gzip, bzip2, and xz are supported out of the box, and zstd is supported if the optional zstandard package is installed.

A clean, uncompressed, plain (or sequence<TAB>count) file streams straight into the C builder in constant memory, which is what makes from_file suitable for very large repertoires; everything else (compressed input, a headered/FASTA/FASTQ file, or a plain file with a byte-order mark, stray carriage returns, or a malformed line near its start) is instead read and validated in Python first, exactly as lzg build does for the same file. One residual case is intentionally not covered by that fast-path check: a malformed line far past the start of an otherwise clean, very large plain file can still reach the C builder and be ingested as-is rather than dropped, since fully validating a multi-gigabyte file up front would defeat the constant-memory streaming from_file exists to provide. For LZGraph.from_file and lzg build, passing strict_input=True / --strict-input validates the whole file up front if you need that stronger guarantee; FlashBackGraph.from_file and lzg flashback build do not currently expose an equivalent flag.

Quick Start: LZGraph

from LZGraphs import LZGraph

# Build a graph from CDR3 amino acid sequences
graph = LZGraph(
    ['CASSLEPSGGTDTQYF', 'CASSDTSGGTDTQYF', 'CASSLEPQTFTDTFFF',
     'CASSLGQGSTEAFF', 'CASSLGIRRT'],
    variant='aap',
)

# Score a sequence
log_p = graph.pgen('CASSLEPSGGTDTQYF')
print(f"log P(gen) = {log_p:.2f}")

# Simulate new sequences
result = graph.simulate(1000, seed=42)
print(f"Generated {len(result)} sequences")

# Diversity
print(f"D(1) = {graph.effective_diversity():.1f}")
print(f"D(2) = {graph.hill_number(2):.1f}")

With gene annotation

from LZGraphs import LZGraph

sequences = ['CASSLEPSGGTDTQYF', 'CASSDTSGGTDTQYF',
             'CASSLEPQTFTDTFFF', 'CASSLGQGSTEAFF']
graph = LZGraph(
    sequences,
    variant='aap',
    v_genes=['TRBV16-1*01', 'TRBV1-1*01', 'TRBV5-1*01', 'TRBV7-2*03'],
    j_genes=['TRBJ1-2*01', 'TRBJ1-5*01', 'TRBJ2-7*01', 'TRBJ1-2*01'],
)

# Gene-constrained simulation
result = graph.simulate(100, sample_genes=True, seed=42)
print(result.v_genes[0], result.j_genes[0])

LZGraph encoding variants

Variant	Input	Node format	Best for
`'aap'`	Amino acid CDR3	`C_2`, `SL_6`	Most TCR/BCR analysis
`'ndp'`	Nucleotide CDR3	`TG0_4`	Nucleotide-level analysis
`'naive'`	Any strings	`C`, `SL`	Motif discovery, ML features

Command line

lzg build repertoire.tsv -o rep.lzg
lzg score rep.lzg sequences.txt
lzg diversity rep.lzg
lzg simulate rep.lzg -n 10000 --seed 42
lzg compare healthy.lzg disease.lzg

Quick Start: FlashBackGraph

from LZGraphs import FlashBackGraph

# Build a Markovian DAG from CDR3 sequences
graph = FlashBackGraph(
    ['CASSLEPSGGTDTQYF', 'CASSDTSGGTDTQYF', 'CASSLEPQTFTDTFFF',
     'CASSLGQGSTEAFF', 'CASSLGIRRT'],
)

# Score a sequence (exact forward DP, no MC)
log_p = graph.pgen('CASSLEPSGGTDTQYF')
print(f"log P(gen) = {log_p:.2f}")

# Simulate from the Markovian distribution
result = graph.simulate(1000, seed=42)

# Diversity, entropy, path count: closed-form via forward DP
print(f"D(1) = {graph.effective_diversity():.1f}")
print(f"D(2) = {graph.hill_number(2):.1f}")
print(f"# distinct paths = {graph.path_count:.3e}")

# SCALE: self-calibrated anomaly score for flagging atypical / error sequences
cal = graph.calibrate_scale(seed=42)               # calibrate once against the graph
print(f"SCALE = {graph.scale_score('CASSLEPSGGTDTQYF', cal):.2f}")  # higher = more anomalous

Build from a file

FlashBackGraph.from_file(path) accepts any of the formats in Input format above (plain, seq<TAB>count, AIRR-style tabular, FASTA, FASTQ, gzip/bzip2/xz-compressed) and is equivalent to running lzg flashback build path -o out.lzg and then loading the result: same detection, same validation, same graph. LZGraph.from_file(path, variant='aap') is the same guarantee for LZGraph, equivalent to lzg build path -o out.lzg.

from LZGraphs import FlashBackGraph

# Write a tiny example file (one CDR3 per line, or seq<TAB>count for abundance)
with open('repertoire.tsv', 'w') as f:
    for s in ['CASSLEPSGGTDTQYF', 'CASSDTSGGTDTQYF', 'CASSLGQGSTEAFF']:
        f.write(s + '\n')

graph = FlashBackGraph.from_file('repertoire.tsv')
print(graph.n_nodes, 'nodes')

For a clean, uncompressed file like this one, from_file streams straight into the C builder in constant memory, so it is the right choice for very large repertoires. For incremental / checkpointed builds over repertoires too large to write to a single file up front, use FlashBackStream: same accumulator with add_sequences(), snapshot(), and finalize(). See the class docstring (help(FlashBackStream)) for the streaming protocol.

When to use which

	`LZGraph`	`FlashBackGraph`
Tokenization	LZ76 dictionary	FlashBack (run-peeling)
Structure	LZ-constrained walks	Markovian DAG
Diversity / entropy / path count	Analytical (with MC where needed)	Closed-form forward DP
Self-calibrated anomaly scoring (SCALE)	No	Yes
V/J gene annotation & gene-conditioned simulation	Yes	No
Encoding variants	`aap`, `ndp`, `naive`	Single representation
CLI tool (`lzg`)	Yes	No
Streaming / incremental build	No	Yes (`FlashBackStream`)

Performance

Benchmark figures below are from a single CPU core on a 5,000-sequence amino-acid CDR3 repertoire (mean length 14.7 aa; resulting LZGraph has ~1,700 nodes, ~9,600 edges). See docs/resources/benchmarks.md for the full table and methodology.

Operation	Throughput
Graph construction	~50,000 sequences/sec (5k seqs in <100 ms)
`pgen()` scoring	~5,000 sequences/sec (constant across batch sizes)
`simulate()`	~4,800 sequences/sec
Hill numbers via MC (10k walks)	~2 sec
Load / save `.lzg`	~100× faster than rebuilding

For repertoires of ~100k sequences and above, graph construction stays linear and saved .lzg files round-trip in seconds. For a clean, uncompressed input file, FlashBackGraph's from_file and FlashBackStream paths operate in bounded memory; we have built and validated graphs with >70,000 nodes and >11M edges this way. (Compressed or headered input is decompressed and validated in memory first, same as lzg build; see Input format.)

Key Capabilities

Every snippet in this section is paste-and-runnable after the Setup block below. graph flags a method that works on either class; lz_graph is an LZGraph instance and fb_graph is a FlashBackGraph instance. Methods marked LZGraph-only or FlashBackGraph-only are not implemented on the other class.

Setup

from LZGraphs import LZGraph, FlashBackGraph, jensen_shannon_divergence

seqs = ['CASSLEPSGGTDTQYF', 'CASSDTSGGTDTQYF', 'CASSLEPQTFTDTFFF',
        'CASSLGQGSTEAFF', 'CASSLGIRRT']
v_genes = ['TRBV5-1*01', 'TRBV5-1*01', 'TRBV5-1*01', 'TRBV7-2*03', 'TRBV7-2*03']
j_genes = ['TRBJ2-7*01', 'TRBJ2-7*01', 'TRBJ2-7*01', 'TRBJ1-2*01', 'TRBJ1-2*01']

lz_graph = LZGraph(seqs, variant='aap', v_genes=v_genes, j_genes=j_genes)
fb_graph = FlashBackGraph(seqs)
graph    = lz_graph                      # `graph` flags methods that work on either class

graph_a      = LZGraph(seqs[:3], variant='aap')
graph_b      = LZGraph(seqs[2:], variant='aap')
population   = LZGraph(seqs * 4, variant='aap')
patient_seqs = ['CASSLGIRRT', 'CASSLGQGSTEAFF']

lz_reference = population
lz_sample    = LZGraph(seqs, variant='aap')

Scoring & Simulation

# Log-probability of a sequence (works on LZGraph and FlashBackGraph alike)
graph.pgen('CASSLEPSGGTDTQYF')               # single → float
graph.pgen(['seq1', 'seq2', 'seq3'])          # batch  → np.ndarray

# Simulate (both classes)
result = graph.simulate(1000, seed=42)
result = lz_graph.simulate(100, v_gene='TRBV5-1*01', j_gene='TRBJ2-7*01')  # LZGraph only

Diversity & Analytics

graph.effective_diversity()          # exp(Shannon entropy)
graph.hill_number(2)                 # inverse Simpson
graph.hill_numbers([0, 1, 2, 5])     # multiple orders → np.ndarray

# LZGraph-only
lz_graph.pgen_distribution()         # analytical log-pgen distribution (Gaussian mixture)
lz_graph.predicted_richness(100_000) # expected unique seqs at depth
lz_graph.predicted_overlap(10000, 50000)        # expected shared sequences
lz_graph.predict_sharing([1000]*5, max_k=5)     # sharing spectrum across donors

# FlashBackGraph-only (closed-form)
fb_graph.path_count                  # exact count of distinct walks
cal = fb_graph.calibrate_scale(seed=0)          # self-calibrate the SCALE anomaly score (once)
fb_graph.scale_score('CASSLEPSGGTDTQYF', cal)   # SCALE: higher = more anomalous
fb_graph.pgen_moments()              # exact moments of log-pgen distribution

Graph Algebra

combined = graph_a | graph_b          # union          (LZGraph and FlashBackGraph)
shared   = graph_a & graph_b          # intersection   (both)
unique_a = graph_a - graph_b          # difference     (both)
personal = population.posterior(patient_seqs, kappa=10.0)  # Bayesian update (both)

Repertoire Comparison

jsd = jensen_shannon_divergence(graph_a, graph_b)  # natural log (nats): 0.0 identical, ln(2) ≈ 0.693 disjoint

ML Feature Extraction

graph.feature_stats()                 # 15-element summary vector (both classes)

# LZGraph-only
lz_reference.feature_aligned(lz_sample)   # project sample into a fixed reference space
lz_graph.feature_mass_profile()           # position-based mass distribution

Serialization

# Both classes use the same .lzg binary format, but each file is class-specific.
lz_graph.save('rep_lz.lzg')
loaded_lz = LZGraph.load('rep_lz.lzg')

fb_graph.save('rep_fb.lzg')
loaded_fb = FlashBackGraph.load('rep_fb.lzg')

Documentation

Full documentation with tutorials, concept guides, and API reference:

https://MuteJester.github.io/LZGraphs/

Quick Start: build your first graph in 5 minutes
Tutorials: graph construction, sequence analysis, diversity metrics
API Reference: complete class and function reference
CLI Reference: terminal tool documentation

Citation

If you use LZGraphs in published research, please cite the methods paper. If you also want to cite a specific software version, add the software entry below.

@article{konstantinovsky2023novel,
  title={A novel approach to T-cell receptor beta chain ({TCRB}) repertoire encoding using lossless string compression},
  author={Konstantinovsky, Thomas and Yaari, Gur},
  journal={Bioinformatics},
  volume={39},
  number={7},
  pages={btad426},
  year={2023},
  publisher={Oxford University Press},
  doi={10.1093/bioinformatics/btad426}
}

@software{lzgraphs_software,
  author={Konstantinovsky, Thomas},
  title={{LZGraphs}: {LZ76} and {FlashBack} compression graphs for immune repertoire analysis},
  url={https://github.com/MuteJester/LZGraphs},
  year={2026}
}

Contributing

Contributions are welcome. Please open an issue or submit a pull request.

Local development setup

LZGraphs builds a CPython extension from a C library at install time, so a working C toolchain is required:

Linux: gcc or clang (any version supporting C11)
macOS: Xcode command-line tools (xcode-select --install)
Windows: Visual Studio Build Tools with the "Desktop development with C++" workload

Then:

git clone https://github.com/MuteJester/LZGraphs.git
cd LZGraphs
pip install -e ".[dev]"   # editable install + dev extras (pytest, pytest-cov, ruff, scipy, build)
pytest                    # run the test suite (~505 tests)

PR checklist

Fork the repository
Create a feature branch (git checkout -b feature/my-feature)
Add tests for new functionality; make sure pytest and pytest tests/regression/ both pass
Commit your changes (small, focused commits preferred)
Push and open a Pull Request describing the motivation and any API changes

License

MIT License. See LICENSE for details.

Contact

Thomas Konstantinovsky, thomaskon90@gmail.com

GitHub · PyPI · Documentation

Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

lzgraphs-3.2.0.tar.gz (449.6 kB view details)

Uploaded Aug 1, 2026 Source

Built Distributions

If you're not sure about the file name format, learn more about wheel file names.

The dropdown lists show the available interpreters, ABIs, and platforms. Enable javascript to be able to filter the list of wheel files.

lzgraphs-3.2.0-cp313-cp313-win_amd64.whl (283.9 kB view details)

Uploaded Aug 1, 2026 CPython 3.13Windows x86-64

lzgraphs-3.2.0-cp313-cp313-musllinux_1_2_x86_64.whl (617.8 kB view details)

Uploaded Aug 1, 2026 CPython 3.13musllinux: musl 1.2+ x86-64

lzgraphs-3.2.0-cp313-cp313-musllinux_1_2_aarch64.whl (613.9 kB view details)

Uploaded Aug 1, 2026 CPython 3.13musllinux: musl 1.2+ ARM64

lzgraphs-3.2.0-cp313-cp313-manylinux_2_24_x86_64.manylinux_2_28_x86_64.whl (627.9 kB view details)

Uploaded Aug 1, 2026 CPython 3.13manylinux: glibc 2.24+ x86-64manylinux: glibc 2.28+ x86-64

lzgraphs-3.2.0-cp313-cp313-manylinux_2_24_aarch64.manylinux_2_28_aarch64.whl (625.2 kB view details)

Uploaded Aug 1, 2026 CPython 3.13manylinux: glibc 2.24+ ARM64manylinux: glibc 2.28+ ARM64

lzgraphs-3.2.0-cp313-cp313-macosx_11_0_arm64.whl (276.2 kB view details)

Uploaded Aug 1, 2026 CPython 3.13macOS 11.0+ ARM64

lzgraphs-3.2.0-cp312-cp312-win_amd64.whl (283.9 kB view details)

Uploaded Aug 1, 2026 CPython 3.12Windows x86-64

lzgraphs-3.2.0-cp312-cp312-musllinux_1_2_x86_64.whl (617.8 kB view details)

Uploaded Aug 1, 2026 CPython 3.12musllinux: musl 1.2+ x86-64

lzgraphs-3.2.0-cp312-cp312-musllinux_1_2_aarch64.whl (613.9 kB view details)

Uploaded Aug 1, 2026 CPython 3.12musllinux: musl 1.2+ ARM64

lzgraphs-3.2.0-cp312-cp312-manylinux_2_24_x86_64.manylinux_2_28_x86_64.whl (627.8 kB view details)

Uploaded Aug 1, 2026 CPython 3.12manylinux: glibc 2.24+ x86-64manylinux: glibc 2.28+ x86-64

lzgraphs-3.2.0-cp312-cp312-manylinux_2_24_aarch64.manylinux_2_28_aarch64.whl (625.1 kB view details)

Uploaded Aug 1, 2026 CPython 3.12manylinux: glibc 2.24+ ARM64manylinux: glibc 2.28+ ARM64

lzgraphs-3.2.0-cp312-cp312-macosx_11_0_arm64.whl (276.2 kB view details)

Uploaded Aug 1, 2026 CPython 3.12macOS 11.0+ ARM64

lzgraphs-3.2.0-cp311-cp311-win_amd64.whl (283.8 kB view details)

Uploaded Aug 1, 2026 CPython 3.11Windows x86-64

lzgraphs-3.2.0-cp311-cp311-musllinux_1_2_x86_64.whl (617.6 kB view details)

Uploaded Aug 1, 2026 CPython 3.11musllinux: musl 1.2+ x86-64

lzgraphs-3.2.0-cp311-cp311-musllinux_1_2_aarch64.whl (614.1 kB view details)

Uploaded Aug 1, 2026 CPython 3.11musllinux: musl 1.2+ ARM64

lzgraphs-3.2.0-cp311-cp311-manylinux_2_24_x86_64.manylinux_2_28_x86_64.whl (627.8 kB view details)

Uploaded Aug 1, 2026 CPython 3.11manylinux: glibc 2.24+ x86-64manylinux: glibc 2.28+ x86-64

lzgraphs-3.2.0-cp311-cp311-manylinux_2_24_aarch64.manylinux_2_28_aarch64.whl (625.3 kB view details)

Uploaded Aug 1, 2026 CPython 3.11manylinux: glibc 2.24+ ARM64manylinux: glibc 2.28+ ARM64

lzgraphs-3.2.0-cp311-cp311-macosx_11_0_arm64.whl (276.5 kB view details)

Uploaded Aug 1, 2026 CPython 3.11macOS 11.0+ ARM64

lzgraphs-3.2.0-cp310-cp310-win_amd64.whl (283.8 kB view details)

Uploaded Aug 1, 2026 CPython 3.10Windows x86-64

lzgraphs-3.2.0-cp310-cp310-musllinux_1_2_x86_64.whl (610.5 kB view details)

Uploaded Aug 1, 2026 CPython 3.10musllinux: musl 1.2+ x86-64

lzgraphs-3.2.0-cp310-cp310-musllinux_1_2_aarch64.whl (606.6 kB view details)

Uploaded Aug 1, 2026 CPython 3.10musllinux: musl 1.2+ ARM64

lzgraphs-3.2.0-cp310-cp310-manylinux_2_24_x86_64.manylinux_2_28_x86_64.whl (620.6 kB view details)

Uploaded Aug 1, 2026 CPython 3.10manylinux: glibc 2.24+ x86-64manylinux: glibc 2.28+ x86-64

lzgraphs-3.2.0-cp310-cp310-manylinux_2_24_aarch64.manylinux_2_28_aarch64.whl (617.2 kB view details)

Uploaded Aug 1, 2026 CPython 3.10manylinux: glibc 2.24+ ARM64manylinux: glibc 2.28+ ARM64

lzgraphs-3.2.0-cp310-cp310-macosx_11_0_arm64.whl (276.5 kB view details)

Uploaded Aug 1, 2026 CPython 3.10macOS 11.0+ ARM64

lzgraphs-3.2.0-cp39-cp39-win_amd64.whl (283.9 kB view details)

Uploaded Aug 1, 2026 CPython 3.9Windows x86-64

lzgraphs-3.2.0-cp39-cp39-musllinux_1_2_x86_64.whl (610.1 kB view details)

Uploaded Aug 1, 2026 CPython 3.9musllinux: musl 1.2+ x86-64

lzgraphs-3.2.0-cp39-cp39-musllinux_1_2_aarch64.whl (606.3 kB view details)

Uploaded Aug 1, 2026 CPython 3.9musllinux: musl 1.2+ ARM64

lzgraphs-3.2.0-cp39-cp39-manylinux_2_24_x86_64.manylinux_2_28_x86_64.whl (620.3 kB view details)

Uploaded Aug 1, 2026 CPython 3.9manylinux: glibc 2.24+ x86-64manylinux: glibc 2.28+ x86-64

lzgraphs-3.2.0-cp39-cp39-manylinux_2_24_aarch64.manylinux_2_28_aarch64.whl (616.9 kB view details)

Uploaded Aug 1, 2026 CPython 3.9manylinux: glibc 2.24+ ARM64manylinux: glibc 2.28+ ARM64

lzgraphs-3.2.0-cp39-cp39-macosx_11_0_arm64.whl (276.5 kB view details)

Uploaded Aug 1, 2026 CPython 3.9macOS 11.0+ ARM64

File details

Details for the file lzgraphs-3.2.0.tar.gz.

File metadata

Download URL: lzgraphs-3.2.0.tar.gz
Upload date: Aug 1, 2026
Size: 449.6 kB
Tags: Source
Uploaded using Trusted Publishing? Yes
Uploaded via: twine/6.1.0 CPython/3.12.9

File hashes

Hashes for lzgraphs-3.2.0.tar.gz
Algorithm	Hash digest
SHA256	`2bf0c59950141b1ed5c12292cc6908ad471226fc6571e6856fa59976b491f886`
MD5	`94c28a883471d477e0cb30279ccc9628`
BLAKE2b-256	`8844bfd966743121911fb6756abdaafcd8362358e37112029b6d84a18b644063`

Algorithm	Hash digest
SHA256	`bf956abbf2b11352031978b396332b5319ad7e090c85dbfbf91bff2992fd1ed5`
MD5	`df60edfc10a216d8e47f98088549b6db`
BLAKE2b-256	`d29f18f276d70791da1257e34149df12fb78d59a578a02efa559678bef409eae`

Algorithm	Hash digest
SHA256	`7a5775deaa93944abb502d2ea88d4a3cb7fd782c7628c334b2ce58bc379bcb9c`
MD5	`53479cdadfac6c5d2885d780960cd2ea`
BLAKE2b-256	`17f7b3613f8ef135d8d5ed4eb98f02fcbcea17e4cca1e3cedf105fd7048be13e`

Algorithm	Hash digest
SHA256	`2a38835c2824884e3667865f493b56c872dcdabd0668a7b55be0a7dd54136636`
MD5	`23d0e8532f79df4d43fdfda5eb0e90a8`
BLAKE2b-256	`99ef8eeb3eea29c49939023ad773a1d40bc06eee3eb274fa2b5ea132ab3df357`

Algorithm	Hash digest
SHA256	`f47fc92a86fc3903c101b0ddf66b563ce6801ac71f2b5e2d454d3f33c18e52b8`
MD5	`992159efe3f48c313cce6455ed73e41f`
BLAKE2b-256	`e3b446fbdf079cd11a78cda7327fef73800620a0f8a097c3a310c9c058e60ec5`

Algorithm	Hash digest
SHA256	`ac133eb53e8df475732d011ad37046eb46cc77bc9b55fc03b1fcdd78b21dd49c`
MD5	`73536e59873581268d0ee5952af6ab90`
BLAKE2b-256	`7eee64e19d1614af934126f5702301e5665c7ac9f2152f5653094ad0af69c50a`

Algorithm	Hash digest
SHA256	`2b6be8307f1cd3aee15f5f06ec90d5c4361f37d2828b57c198f94cb0eda782ac`
MD5	`e55f8a60140c5e2d5b9ba72193eb956d`
BLAKE2b-256	`6d82f2a7da0c2a5fb15129a8bcbe58dbb944acad74a01598052017a8be9d16cc`

Algorithm	Hash digest
SHA256	`fbd6b98e71c6fb69d4df01c79253faa5ecdcdcbcc7729e61442cdc4eb6287e8a`
MD5	`89e12595adbbb23405bcd289d5595e08`
BLAKE2b-256	`922822d2a6c7c3a15cbabda0373f3f7750f0890f6b5d19d638e664055a83224b`

Algorithm	Hash digest
SHA256	`28d9d31105c599f07c81516622ec0bfe7a8c6bf39303ee906491b80e1bebf5b3`
MD5	`b87a75b5e0cb96ffa658e042572e292f`
BLAKE2b-256	`5022854d509624e98e809721b6cbc2d014ea78643f442542c9729de36fa3bb11`

Algorithm	Hash digest
SHA256	`183e414cb278ad02b01a3101f7a6a8366b3b6d524cfe2c27737ae03ef04f2a23`
MD5	`5bcd2fc13ae941c561d010f9bec2264b`
BLAKE2b-256	`5eb7f9ad2c7825b41fe7c863fe8898cdb1782d457fb058c8f81f7ac5cb56268c`

Algorithm	Hash digest
SHA256	`d8a9c1cd3196447abc6c4bceff85bf8c5476fded5155bbfed563b6b51e7841d8`
MD5	`0296d89353f0a5616fccc10c11655e17`
BLAKE2b-256	`670ffc50f3205aeaa3dda1763428dc133f295b35f84ead5b55502ca42e48180c`

Algorithm	Hash digest
SHA256	`8f004983137722d0fb6ff991d1ef2e07563ec45aec7c19ab4e41b6a4cacf76f1`
MD5	`b4fb907f2cece2ff26c8eba96cb39857`
BLAKE2b-256	`810f074cdc88ced608934ec7ff9838ab3978d904665e5778fb4aa3b6476b087c`

Algorithm	Hash digest
SHA256	`d1a49af9e37f20233cfda901c03d605098974f149efbf292115487bc8a7699e2`
MD5	`ca03e8727fd5c998b00288fe3dba5d84`
BLAKE2b-256	`7616602d5f4e1b0a78bdb725a4c0c4b1fceb2b4e926bd071bfd4637fd66b1625`

Algorithm	Hash digest
SHA256	`2ad51a5c95cdd90a54e3ec99cbb33c694a845be94315b0bc5e23c28cb041f265`
MD5	`8b82f7dc004a06fa0e4faf37363a1659`
BLAKE2b-256	`2ca582fc9f0ae39dbc575f1e829be13030d71f91d270db4e9f2d8f5ef477b465`

Algorithm	Hash digest
SHA256	`59799ec235b4bbed927c413b4600e7bc331c1385e98fd27aa61a77874211eb09`
MD5	`be094df6d893a29a2f7645a9f868bd55`
BLAKE2b-256	`b16cd8b482b105a284b03cb34487bd242b0c8956d04ab3467e120c28edc58d76`

Algorithm	Hash digest
SHA256	`a61147bde22b385dd7856840a69610e364ba2f977a9a346da2c0193f01eb2cd7`
MD5	`b1f5f1dcbd343aed8f5a0bff5366ece5`
BLAKE2b-256	`0c2a3ffb87296a60cbdb9fcc64af09504001c4ed3bb5b74fbdcc7e0b840fa8c2`

Algorithm	Hash digest
SHA256	`cf3fa0fb4d6d10c1ccf82cbd3c1f55d4358630bc30a804095c6b5d31a3a6786b`
MD5	`426faebbf76581655746fa07495f83d5`
BLAKE2b-256	`10f604e20549b07bee3d6f1363c8b326fd1447d154f5d247b5030319b5975794`

Algorithm	Hash digest
SHA256	`4f754f8c186e7e5edb6c9ac56a8a0d63485be811b1437248f6d305c08bd4f9cc`
MD5	`3ec0f24c8c56fe31b12f0f0ec7e02275`
BLAKE2b-256	`98389aa3055bf1df699d1538af1b84fbcfa0f96ceb974f2b199cd0dc639e3987`

Algorithm	Hash digest
SHA256	`a35fbaa42256b2284304eb17ede0eca47feba5fe8797bc2a6901ef10041c91f4`
MD5	`9d4ca8ebd955fd2a6433aededfa8ba42`
BLAKE2b-256	`526cd84fa08d01a2785521fcfb328e2de6133a2ff30a76280ddd55a7c0c2114a`

Algorithm	Hash digest
SHA256	`98a656ade5fa537c484dc860e8f490c41e679952ae598d8a736df01721c197a8`
MD5	`57dcaaa6484015f2bcf2b82acf32e9f8`
BLAKE2b-256	`b21c798bd08703d563011c5a45256e4a1ed4ba6284167ed71af1677efcf433b2`

Algorithm	Hash digest
SHA256	`315cc7cd140913fdb72d6e494a3a15b96c4103601468c4785a0be33f8cf5bb38`
MD5	`9d5c8354f8b2aa9283edcba56cba2c4c`
BLAKE2b-256	`78a685f7edfbe150a3ad6badf64582299bd49efa6e02d82157f53768d2b9cdc1`

LZGraphs 3.2.0

Installation

Input format

Quick Start: LZGraph

With gene annotation

LZGraph encoding variants

Command line

Quick Start: FlashBackGraph

Build from a file

When to use which

Performance

Key Capabilities

Setup

Scoring & Simulation

Diversity & Analytics

Graph Algebra

Repertoire Comparison

ML Feature Extraction

Serialization

Documentation

Citation

Contributing

Local development setup

PR checklist

License

Contact

Project links

Key dates

1 maintainer

Credits

GitHub Statistics

License expression

Requires

Provides Extra

Tags

Classifiers

Download files

Source Distribution

Built Distributions

File details

File metadata

File hashes

Provenance

File details

File metadata

File hashes

Provenance

File details

File metadata

File hashes

Provenance

File details

File metadata

File hashes

Provenance

File details

File metadata

File hashes

Provenance

File details

File metadata

File hashes

Provenance

File details

File metadata

File hashes

Provenance

File details

File metadata

File hashes

Provenance

File details

File metadata

File hashes

Provenance

File details

File metadata

File hashes

Provenance

File details

Algorithm	Hash digest
SHA256	`39f6e2774f5eec1e12d8c1429c16dd403f25bd7f0ddb3c2a137bd8e77ebfbbd9`
MD5	`5bc1bc68963c63eb4704ee8c71450804`
BLAKE2b-256	`b5bb69e7a78c75f25065f47c0f9ae8ab5687e55f0bc09319d700f8d462074986`

Algorithm	Hash digest
SHA256	`1f377a2f39f5618cf7696c44e808493f74fe0284456296fb3e2d982252d5ea05`
MD5	`aa5993b22dac6b1cc0d3b71d6a7444d7`
BLAKE2b-256	`3a4fb518d0f3b118fd0b68dd9b560650fe96c4cdb9751e3ca094f15fbeacef68`

Algorithm	Hash digest
SHA256	`2678faaeaf2be27392a2bdf4bfdb73c7b75a2b5e996c3bf04108b29983907347`
MD5	`d000f0f8923fb71b8e870c5bfec91d94`
BLAKE2b-256	`db05fc4d469ae5a2ef03b3d74ca0ded93b2d20a3ef2fda1f0b273476dcb02c26`

Algorithm	Hash digest
SHA256	`84b49caa271c62dbe481677cba80a1ad849d478ad7c63410b26c06b7e9580605`
MD5	`a1e2c40af5b478c8fad76b2039de36f7`
BLAKE2b-256	`5e5d6dda58a9821f47ff7e0a180e892edb9c227740c94a7e6593533b2f1a08ea`

Algorithm	Hash digest
SHA256	`3dd1f97d52a5d91ae3d95e9fe3660e9e1724b31fc0c2165c8a3d1d27660a3f80`
MD5	`1749b2bb8deb5d5ff66e9302f4e465e4`
BLAKE2b-256	`51b37edf6ebc05100b3dc1527d4f6c9830e2f3cc52fa5278ef812ba0b2e25183`

Algorithm	Hash digest
SHA256	`28567f9c069204b7dde95a4e2cd8086a1e1f30c7b00121744e0293c7d5a22614`
MD5	`209c51810e1bbd9c8555c84d0b0d36a1`
BLAKE2b-256	`1131ecc61f976f222fc30ab4cdfb64dbbce428e3c3dda9e1e1600f5ea554d6f1`

Algorithm	Hash digest
SHA256	`964232e84201b112b3566dd6155da7cd9cdf7d04478f5e82a0f30177bc74d929`
MD5	`93425daecaa1158c0bfd47e0502cc8d1`
BLAKE2b-256	`368f39b64ffb8ce903248c4159b79da28faf89ad01c6d5be4c22ce0ff4071236`

Algorithm	Hash digest
SHA256	`fe899dcd3e3552f8987b1ecdac26b09b8b14488feca4a862da9d6333f726744a`
MD5	`6e7670237eb5a543c6c9f52d1661296f`
BLAKE2b-256	`dda3ed2086d7c6f7da67360a5bc17b10584920ce5d6e1b0f9ff71a53838e89f8`

Algorithm	Hash digest
SHA256	`95ce6c8986649ad3d9da41f32ff8b777302eae20476aa64717c0af48817995af`
MD5	`300d9f5d543babd63f18ba914c616340`
BLAKE2b-256	`2987c0d2e8a386a37550a9a3a36bf12df98d8dede0142622ac3917cf1089c982`

Algorithm	Hash digest
SHA256	`cd7915690b06c90588cf8396b8a52a6b9e24b3a5c2cbdf0cac420eba6136a4fc`
MD5	`cbac6449aaddacd8dbd494d9a22b069f`
BLAKE2b-256	`9ddd17d0b8185d3a01006c866ce4c6487d605fa186f954a72b5edddf096f1b54`