eFISHent

With BLAST+ and transcriptome tools (for transcriptome-level off-target filtering):

curl -LsSf https://raw.githubusercontent.com/BBQuercus/eFISHent/main/install.sh | bash -s -- --with-blast

Custom install path:

curl -LsSf https://raw.githubusercontent.com/BBQuercus/eFISHent/main/install.sh | bash -s -- --prefix /path/to/install

Update:

efishent --update

Development install:

git clone https://github.com/BBQuercus/eFISHent.git
cd eFISHent/
./install.sh --deps-only
uv venv && source .venv/bin/activate
uv pip install -e .

Uninstall:

curl -LsSf https://raw.githubusercontent.com/BBQuercus/eFISHent/main/install.sh | bash -s -- --uninstall

Or simply: rm -rf ~/.local/efishent

For any organism, download the genome FASTA and GTF annotation from Ensembl or UCSC. Prefer primary_assembly if available, otherwise toplevel. Unzip with gunzip.

Example for human (GRCh38):

# Reference genome
wget https://ftp.ensembl.org/pub/current_fasta/homo_sapiens/dna/Homo_sapiens.GRCh38.dna.primary_assembly.fa.gz
gunzip Homo_sapiens.GRCh38.dna.primary_assembly.fa.gz

# GTF annotation (for intergenic filtering, rRNA filtering, expression weighting)
wget https://ftp.ensembl.org/pub/current_gtf/homo_sapiens/Homo_sapiens.GRCh38.115.gtf.gz
gunzip Homo_sapiens.GRCh38.115.gtf.gz

Ensembl GTFs use gene_biotype while GENCODE uses gene_type — eFISHent supports both.

Reference transcriptome (optional, for BLAST cross-validation):

gffread Homo_sapiens.GRCh38.115.gtf -g Homo_sapiens.GRCh38.dna.primary_assembly.fa -w transcriptome.fa

# Append rRNA sequences (18S/28S/5.8S are NOT in standard GTFs)
efetch -db nucleotide -id NR_003286.4 -format fasta >> transcriptome.fa  # 45S pre-rRNA
efetch -db nucleotide -id NR_023363.1 -format fasta >> transcriptome.fa  # 5S rRNA

The major rRNA genes exist in ~300 tandem copies in unassembled regions, so they’re absent from standard GTFs. Including them in the transcriptome FASTA ensures the BLAST filter catches probes binding these abundant sequences.

Count table (optional, for expression-weighted filtering):

Download a normalized RNA-seq dataset (FPKM/TPM) for your cell line from GEO or Expression Atlas. The file needs Ensembl gene IDs in column 1 and normalized counts in column 2.

flowchart TD
    A["Gene Sequence
FASTA file or NCBI download"] --> B["Generate Candidate Probes
Sliding window (adaptive or fixed length)"]

    B --> C["Basic Filtering
TM, GC, homopolymers, low-complexity"]

    C --> D["Genome Alignment
Bowtie2 (default) or Bowtie
+ repeat masking, intergenic, Tm scoring"]

    D --> E{"Transcriptome
provided?"}
    E -- Yes --> F["Transcriptome BLAST
Off-target detection (TrueProbes params)"]
    E -- No --> G["K-mer Filtering
Jellyfish frequency count"]
    F --> G

    G --> H["Secondary Structure
deltaG prediction (RNAstructure)"]

    H --> H2["Accessibility Scoring
RNA folding (optional)"]

    H2 --> I["Quality-Weighted Optimization
Greedy or optimal (MILP) + gap filling + Tm refinement"]

    I --> K["Validation Report
Quality scores, off-target genes, recommendations"]

    K --> J["Final Probe Set"]

    style A fill:#e1f5fe
    style J fill:#e8f5e9

1. Candidate probes are generated from the input sequence using a sliding window. When --adaptive-length is enabled, probe lengths are adjusted based on local GC content to normalize Tm.
2. Basic filtering removes probes failing sequence criteria: melting temperature, GC content, homopolymer runs, optionally low-complexity regions, and optionally G-quadruplex motifs (--filter-g-quadruplex).
3. Probes are aligned to the reference genome using Bowtie2 (sensitive local alignment with OligoMiner/Tigerfish parameters). Optional filters refine off-target counting: repeat masking, intergenic filtering, thermodynamic scoring, and expression weighting.
4. If a reference transcriptome is provided, probes are BLASTed against expressed transcripts to catch off-targets that genome alignment alone may miss (e.g., splice junctions).
5. Short k-mers are counted using Jellyfish — probes with frequently occurring k-mers are discarded.
6. Secondary structure is predicted using a nearest-neighbor thermodynamic model — probes with too-stable structures are filtered.
7. If --accessibility-scoring is enabled, target RNA accessibility is scored using RNA folding predictions.
8. Quality-weighted optimization selects non-overlapping probes maximizing coverage. A gap-filling pass covers remaining regions and Tm uniformity refinement swaps outlier probes.
9. The output includes per-probe quality scores, off-target gene names, expression risk, and PASS/FLAG/FAIL recommendations.

Genome alignment (default):

Transcriptome BLAST (optional):

Expression weighting (optional):

smFISH with pre-built index (simplest):

efishent --genome hg38 --gene-name "TP53" --organism-name "homo sapiens" --preset smfish --threads 8

smFISH with full off-target filtering:

efishent \
    --reference-genome ./hg-38.fa \
    --reference-annotation ./hg-38.gtf \
    --reference-transcriptome ./transcriptome.fa \
    --gene-name "GAPDH" \
    --organism-name "homo sapiens" \
    --preset smfish \
    --mask-repeats True \
    --intergenic-off-targets True \
    --filter-rrna True \
    --max-probes-per-off-target 5 \
    --threads 8

Long probes (45-50nt) with optimal solver:

efishent \
    --reference-genome ./hg-38.fa \
    --gene-name "norad" \
    --organism-name "homo sapiens" \
    --is-plus-strand True \
    --optimization-method optimal \
    --min-length 45 \
    --max-length 50 \
    --formamide-concentration 45 \
    --threads 8

Exogenous gene (GFP, Renilla, etc.):

efishent \
    --reference-genome ./hg38.fa \
    --reference-transcriptome ./transcriptome.fa \
    --reference-annotation ./hg38.gtf \
    --sequence-file "./renilla.fasta" \
    --preset exogenous \
    --threads 8

Expression-weighted off-target filtering:

efishent \
    --reference-genome ./hg-38.fa \
    --reference-annotation ./hg-38.gtf \
    --ensembl-id ENSG00000128272 \
    --organism-name "homo sapiens" \
    --is-plus-strand False \
    --max-off-targets 5 \
    --encode-count-table ./count_table.tsv \
    --max-expression-percentage 20 \
    --threads 8

Rescue probes with thermodynamic and repeat masking filters:

efishent \
    --reference-genome ./hg-38.fa \
    --reference-annotation ./hg-38.gtf \
    --sequence-file ./my_gene.fasta \
    --mask-repeats True \
    --intergenic-off-targets True \
    --off-target-min-tm 37 \
    --threads 8