LncNova

Package
Meta
Author	AIIMS New Delhi

An integrated RNA-seq pipeline for novel lncRNA discovery and characterization.

LncNova takes raw RNA-seq FASTQ files and returns a curated set of novel lncRNAs (transcripts absent from GENCODE and LNCipedia), then characterizes them. It runs on a single sample or on a case/control cohort. All tools and reference databases ship in a container image, so nothing else needs to be installed.

FASTQ ─ fastp ─ HISAT2 ─ strandedness check ─ StringTie (per sample) ─ merge + gffcompare
      ─ novelty / length / exon-count / TPM filters ─ CPC2 + CPAT consensus ─► novel lncRNAs
      ─ featureCounts (known genes vs GENCODE, novel genes vs the novel set)

cohort only:  PyDESeq2 ─ cis / trans targets ─ GO / KEGG ─ IntaRNA (lncRNA + mRNA 3'UTR) ─ ceRNA triads
all modes:    cis targets ─ IntaRNA miRNA sites ─ disease precedent (RNADisease)

Install

LncNova is a thin Python launcher plus a container image. You need Python ≥ 3.9 and one of Docker, Apptainer or Singularity.

pip install lncnova                                        # the `lncnova` command (launcher only)

The PyPI package does not contain the tools or databases. They live in the container image, which you build once from the Dockerfile in the source repository:

git clone <this-repository> && cd LncNova
docker build -t lncnova:0.1.0 --build-arg THREADS=16 .    # downloads and bakes in all data (~20 GB image)
lncnova check                                              # verifies every tool and database in the image

The image build downloads GENCODE v49, builds a HISAT2 index, and fetches LNCipedia 5.2, miRBase, the CPAT human models, Enrichr GO-BP/KEGG gene sets and RNADisease v4.0. It needs a good connection and roughly an hour. docker/README.md in the repository explains running the image directly, on HPC with Apptainer, or offline.

Run

lncnova template > samples.tsv        # edit the paths
lncnova run -s samples.tsv -o results -t 16

samples.tsv (TSV or CSV), one row per sample:

Paths are bind-mounted into the container automatically. Re-running resumes: finished per-sample steps are skipped, and merge/quantification are recomputed if the sample set changes.

Options

Output

results/
├── lncnova_report.xlsx        summary + every table below
├── 06_novel_lncrna/           novel_lncrna.gtf, novel_lncrna.fa, novel_lncrna_transcripts.tsv
├── 05_quant/                  transcript_tpm.csv, gene_counts.csv
├── 07_de/                     DESeq2-style tables for all, novel, known-lncRNA and coding genes (cohort)
├── 08_targets/                cis_targets.csv, trans_targets.csv (cohort)
├── 09_mirna/                  IntaRNA hits and ceRNA triads
├── 10_disease_precedent.csv
└── 01_trim … 04_merge/        per-step files and logs

How the calls are made

• Novel: gffcompare class u (intergenic), x (antisense) or i (intronic) against GENCODE v49, length ≥ 200 nt, ≥ 2 exons, no overlap with a same-strand LNCipedia 5.2 exon, and TPM ≥ 0.5 in enough samples.
• Non-coding: CPC2 and CPAT must both call the transcript non-coding (CPAT human cutoff 0.364).
• Counts: known genes are counted against GENCODE and novel genes against the final novel set. StringTie's merged annotation relabels many known genes as MSTRG, so it cannot supply both.
• Co-expression: Pearson on log-normalized counts after removing group means, Benjamini-Hochberg adjusted, |r| ≥ 0.6. It is skipped with fewer than 6 samples.
• miRNA sites: IntaRNA, seed of 7 bp enforced, energy ≤ cutoff. lncRNAs are tested as whole transcripts; mRNAs by their 3′UTR, and only against miRNAs that bind a lncRNA.
• Disease precedent: each cis or trans target gene is looked up in RNADisease v4.0.

Limits

• Human GRCh38 / GENCODE v49 only; paired- or single-end FASTQ.
• Differential expression uses PyDESeq2, which is close to but not identical to R DESeq2.
• HISAT2 uses a plain genome index plus the known splice sites at alignment time. The splice-graph index needs far more memory than most machines have.
• Co-expression and ceRNA results are hypothesis-generating and are only convincing with roughly 10 or more samples. The ceRNA and disease layers are computational associations, not evidence for a specific lncRNA. Validate candidates experimentally.
• Tested end to end on subsampled data (4 and 6 samples, 2 million read pairs each). Runtime and results on a full-depth cohort have not been benchmarked.

Development

pip install -e ".[test]" && pytest
python -m build            # wheel and sdist in dist/

License

LncNova's code is MIT-licensed (see LICENSE.txt). The image bundles third-party tools and data under their own terms: HISAT2, Subread and CPAT are GPL; gffcompare is Artistic 2.0; IntaRNA is MIT but uses ViennaRNA, whose license should be checked before commercial redistribution of the image; GENCODE, LNCipedia, miRBase, Enrichr and RNADisease data follow their providers' terms.