Changelog

v1.3.2.5 — 2026-10

Fixes

cis-gs extract and cis-gs batch now reverse-complement minus-strand promoters with the full IUPAC complement table. Previously only A, C, G, T and N were complemented and any other ambiguity code (R, Y, K, M, B, D, H, V) was left unchanged. The effect is limited to promoters that contain ambiguity codes (1 of 33,623 promoters in the rice IRGSP-1.0 RefSeq genome); all other promoters were already identical to bedtools and seqkit output. The GUI was not affected.
cis-gs --help no longer crashes on consoles that use the Windows cp1252 code page (a non-ASCII arrow in one help string); the CLI now substitutes any character the console cannot print.

Fixes

cis-gs extract and cis-gs batch no longer fail with 'str' object has no attribute 'open' (path arguments are now coerced to Path).
Motif significance model corrected. The per-promoter binomial test now uses each promoter’s own base composition (previously a single GC value estimated from the hit sequences themselves), and the Benjamini-Hochberg correction is applied over all promoters x motif strands scanned (previously only over pairs that already had a hit, which flagged nearly every hit as significant). The set of hits is unchanged; the p_value, p_value_adj and significance columns change, so results from earlier versions should not be compared directly. Validated on composition-preserving shuffled promoter sets (no false positives).

Features

cis-gs search now reports a motif-level enrichment summary (observed versus expected hits under the composition-matched null) and the number of hits significant at FDR <= 0.05.

Performance

Promoter extraction reads through a FASTA offset index (samtools .fai compatible, built in memory if absent) instead of loading whole chromosomes. On the 2.6 Gb Arachis hypogaea genome (106,608 promoters) extraction dropped from ~63 s / 1.9 GB to ~5 s / 0.4 GB with byte-identical output. The GUI build previously re-read the chromosome for every gene and benefits as well. Irregular FASTA layouts fall back to the previous reader.
Motif scanning parses the FASTA once instead of once per motif.

Fixes

Fixed a severe performance bug in the Step 2 motif-search significance (p-value / FDR) calculation: the per-gene/motif p-value lookup re-scanned the entire hits table on every iteration (an accidental O(n²) pattern), which could hang indefinitely on genome-wide promoter sets with a short, common motif (e.g. a 5 bp IUPAC pattern against 100,000+ promoters). Replaced with a precomputed lookup dict and vectorized pandas indexing. Verified on a 106,608-promoter genome-wide run: full scan + significance calculation now completes in ~21 seconds (previously did not finish after 30+ minutes).

Fixes

Cleaned up the authors / maintainers fields in pyproject.toml so that the package’s Author metadata renders cleanly as Ayushman Mallick on PyPI and pepy.tech. The Plant Signaling Lab, IISER Tirupati attribution is now carried in the Maintainer field, eliminating comma-driven truncation issues on third-party indexes.

Fixes

Replaced broken shields.io pypi/dt downloads badge with the pepy.tech all-time downloads badge in the README, ensuring the downloads counter renders correctly on both GitHub and PyPI.

Fixes

Corrected PyPI author metadata — Author field now correctly displays the authors on PyPI and pypistats.org.

Fixes

Highlights

Multi-species batch mode — new cis-gs batch subcommand accepts a tab-separated manifest file (species, FASTA path, GFF3 path, optional upstream bp) and runs promoter extraction + motif search for every species in a single automated pass, writing per-species hit CSVs and a combined batch_hits.csv with a species column.
Batch wizard — cis-gs wizard batch provides a step-by-step interactive guide for the new batch workflow, consistent with all other wizard commands.

Documentation

Added cis-gs batch to the CLI command reference.
Rewrote the Wizard page with a full workflow table and dedicated multi-species batch section.
Added missing cis-gs logo command to the CLI reference (was implemented but not documented).

Highlights

37× faster Louvain clustering — replaced the O(N²) Python double loop for co-expression graph construction with NumPy vectorised masking (np.triu), and switched the community-detection backend from python-louvain + NetworkX to the igraph C-library (community_multilevel). Louvain time for A. hypogaea (56 513 genes) dropped from ~150 s to ~4 s.
igraph added as a dependency — igraph >= 0.11 added to pyproject.toml; python-louvain retained as fallback.
PyInstaller bundle updated — igraph and igraph._igraph added to hiddenimports in Cis-GS.spec; Windows .exe rebuilt and uploaded to the v1.2.0 GitHub release.

Performance (benchmarked)

Promoter extraction: 11–87 s across four genome assemblies (O(C) disk reads, where C = chromosome count).
Motif scanning: 6–18 s scaling linearly with sequence count.
KEGG REST retrieval: 5–8 s on first use; < 30 ms on subsequent calls via local disk cache (229–311× speedup).
Full co-expression pipeline (normalisation → Pearson → Louvain → k-means): 77–159 s per species.

Fixes

All em dashes replaced with hyphens throughout GUI and CLI output.
scan_fasta_for_motifs import aligned with the canonical function name in cis_gs.app.

Highlights

Live KEGG organism dropdown (11 700+ organisms, fetched via REST).
Live NCBI Taxonomy search in both GUI and CLI.
60× faster MyGene.info ID conversion via batched POST.
Interactive CLI wizard plus -i / --interactive on every command.
Fuzzy “did you mean…?” for CLI typos.
Modern single-colour theme (#16A085 teal) with instant theme swap (pre-cached stylesheet, setUpdatesEnabled freeze).
First-run NCBI email prompt.
Three Gene-ID-Mapping methods (column swap, mapping CSV, GFF3 Dbxref expansion).
Real brand-icon Contact tab.

Removed

Fixes

Initial public release.

Seven-step pipeline: Promoters → Motifs → Logos → Expression → Coexpression → K-means → Enrichment.
PlantTFDB importer.
PyQt5 GUI + first-cut CLI.
Standalone Windows .exe via PyInstaller.