Skip to content

Which Graph Should I Use?

LZGraphs has two graph families. Pick by the question you are asking, not by the algorithm underneath.

Quick answer

If you want to... Use
Exact diversity, Hill numbers, or entropy with no sampling noise FlashBackGraph
Self-calibrated per-sequence anomaly scores (SCALE) FlashBackGraph
Exact generation probability and the most or least probable sequences FlashBackGraph
V/J gene-aware modeling LZGraph
ML feature vectors (reference-aligned, mass profiles, stats) LZGraph
Occupancy and sharing predictions (richness at depth, sample overlap) LZGraph
Very large repertoires with constant-memory streaming construction LZGraph

The difference in one paragraph

LZGraph uses a coarsened LZ76 dictionary. It is the general-purpose tool: it carries V/J gene statistics, scales to large repertoires, produces ML features, and answers occupancy and sharing questions. Diversity and generation are estimated efficiently, with simulation where a closed form is intractable.

FlashBackGraph builds a strictly Markovian graph, so diversity, entropy, Hill numbers, and generation probability are computed exactly by forward dynamic programming (no sampling), and it provides the fast FlashBack Anomaly Score for surprise detection. It does not model V/J genes.

Still unsure?

Start with LZGraph (variant='aap'). It covers the widest range of repertoire analysis tasks. Move to FlashBackGraph when you specifically need exact, sampling-free analytics or anomaly scoring.

Next steps