bio-genome-annotation-ncrna-annotation

name	bio-genome-annotation-ncrna-annotation
description	Identify non-coding RNAs including tRNAs, rRNAs, snoRNAs, and regulatory RNAs using Infernal covariance model searches against Rfam and tRNAscan-SE for tRNA prediction. Use when performing genome-wide ncRNA annotation with assembly input producing GFF output.
tool_type	cli
primary_tool	Infernal

Non-Coding RNA Annotation

Identify and annotate non-coding RNAs in genome assemblies using Infernal (general ncRNAs via Rfam covariance models) and tRNAscan-SE (specialized tRNA detection).

Infernal / cmscan

Infernal uses covariance models (CMs) from Rfam to identify ncRNA families by both sequence and secondary structure similarity.

Rfam Database Setup

# Download Rfam covariance models (~600 MB compressed)
wget https://ftp.ebi.ac.uk/pub/databases/Rfam/CURRENT/Rfam.cm.gz
gunzip Rfam.cm.gz

# Press the CM database (required for cmscan)
cmpress Rfam.cm

# Download clan information (for overlap resolution)
wget https://ftp.ebi.ac.uk/pub/databases/Rfam/CURRENT/Rfam.clanin

Basic cmscan

# Search genome against all Rfam families
# --cut_ga: Use Rfam gathering threshold (recommended, family-specific)
# --rfam: Run in Rfam mode (skip slow alignment for large models)
# --nohmmonly: Use CM scoring only (more accurate, slower)
# --fmt 2: Output in table format
cmscan \
    --cut_ga \
    --rfam \
    --nohmmonly \
    --tblout ncrna_hits.tbl \
    --fmt 2 \
    --cpu 16 \
    --clanin Rfam.clanin \
    Rfam.cm \
    genome.fasta > ncrna_hits.out

Key Options

Option	Description
--cut_ga	Use Rfam gathering threshold (recommended)
--rfam	Speed optimization for Rfam-scale searches
--nohmmonly	Force CM-only scoring (more sensitive for structured RNAs)
--clanin	Rfam clan file for resolving overlapping hits
--tblout	Tabular output file
--fmt 2	Extended table format with accessions
--cpu	CPU threads
-E	E-value threshold (default: 10; --cut_ga is preferred)

Convert cmscan Output to GFF3

# Using Infernal's built-in conversion
grep -v '^#' ncrna_hits.tbl | \
awk 'BEGIN{OFS="\t"} {
    if ($10 == "+") strand = "+"; else strand = "-";
    if ($10 == "+") {start=$8; end=$9} else {start=$9; end=$8};
    print $4, "Infernal", "ncRNA", start, end, $17, strand, ".", "ID="$2";Name="$3";rfam_acc="$2";evalue="$17
}' > ncrna_infernal.gff3

E-value Guidelines

Category	Typical E-value	Notes
High confidence	< 1e-10	Unambiguous family assignment
Moderate	1e-10 to 1e-5	Likely real, verify context
Rfam GA threshold	family-specific	Recommended cutoff with --cut_ga
Marginal	> 1e-3	May be pseudogenes or degraded copies

tRNAscan-SE

Specialized tRNA detector using covariance models. More sensitive and specific for tRNAs than Infernal alone.

Basic Usage

# Bacterial/archaeal mode
tRNAscan-SE -B -o trna_results.txt --gff trna.gff3 genome.fasta

# Eukaryotic mode
tRNAscan-SE -E -o trna_results.txt --gff trna.gff3 genome.fasta

# General mode (auto-detect domain)
tRNAscan-SE -G -o trna_results.txt --gff trna.gff3 genome.fasta

Key Options

Option	Description
-B	Bacterial mode
-A	Archaeal mode
-E	Eukaryotic mode
-G	General (mixed/unknown) mode
-o	Tabular output
--gff	GFF3 output
--detail	Detailed output with isotype info
--thread	CPU threads
-Q	Covariance model scoring only (slower, more accurate)

Domain-Specific Models

Domain	Flag	Notes
Bacteria	-B	Shorter introns, smaller genomes
Archaea	-A	Split tRNAs, bulge-helix-bulge introns
Eukaryota	-E	Longer introns, pseudogenes common
Organelle	-O	Mitochondrial/chloroplast tRNAs
Mitochondrial	-M	Mammalian mitochondrial-specific

Expected tRNA Counts

Organism Type	Expected tRNAs	Notes
Bacteria	30-90	Fewer isotypes, some shared
Archaea	30-60	Similar to bacteria
Yeast	275-400	Many isodecoders
Nematode	600-900	Gene family expansion
Mammal	400-600	Many pseudogenes
Plant	500-1,000+	Large gene families

barrnap (rRNA Detection)

barrnap predicts rRNA genes using HMM models. Note: barrnap has been unmaintained since 2018. Bakta handles rRNA detection internally for prokaryotes.

# Detect rRNAs
barrnap --kingdom bac genome.fasta > rrna.gff3

# Kingdoms: bac (bacteria), arc (archaea), euk (eukaryote), mito (mitochondria)
barrnap --kingdom euk --threads 4 genome.fasta > rrna.gff3

Combining ncRNA Annotations

import pandas as pd
from collections import defaultdict

def parse_infernal_tbl(tbl_file):
    '''Parse Infernal cmscan tabular output.'''
    hits = []
    with open(tbl_file) as f:
        for line in f:
            if line.startswith('#'):
                continue
            parts = line.split()
            if len(parts) < 18:
                continue
            strand = '+' if parts[9] == '+' else '-'
            start, end = (int(parts[7]), int(parts[8])) if strand == '+' else (int(parts[8]), int(parts[7]))
            hits.append({
                'target': parts[0],
                'rfam_acc': parts[1],
                'rfam_name': parts[2],
                'seqid': parts[3],
                'start': start,
                'end': end,
                'strand': strand,
                'evalue': (parts[]),
                : (parts[]),
                : classify_rfam(parts[]),
            })
     pd.DataFrame(hits)

 ():
    
    name_lower = rfam_name.lower()
       name_lower    name_lower    name_lower:
         
       name_lower:
         
       name_lower    name_lower    name_lower:
         
       name_lower:
         
       name_lower    name_lower:
         
       name_lower    name_lower:
         
     

 ():
    
    trnas = []
     (gff_file)  f:
         line  f:
             line.startswith():
                
            parts = line.strip().split()
             (parts) <   parts[] != :
                
            attrs = (item.split()  item  parts[].split()    item)
            trnas.append({
                : parts[],
                : (parts[]),
                : (parts[]),
                : parts[],
                : attrs.get(, ),
                : attrs.get(, ),
                : (parts[])  parts[] !=   ,
                : ,
            })
     pd.DataFrame(trnas)

 ():
    
    infernal_df = parse_infernal_tbl(infernal_tbl)
    trna_df = parse_trnascan_gff(trnascan_gff)

    
    infernal_no_trna = infernal_df[infernal_df[] != ]

    summary = defaultdict()
     rna_type  infernal_no_trna[].unique():
        summary[rna_type] = (infernal_no_trna[infernal_no_trna[] == rna_type])
    summary[] = (trna_df)

    ()
     rna_type, count  (summary.items()):
        ()
    ()

     infernal_no_trna, trna_df, summary

Troubleshooting

cmscan Is Slow

• Use --rfam flag for Rfam-mode optimizations
• Use --FZ 2 to further speed up (less sensitive)
• Split genome into chunks and run in parallel

Missing Expected ncRNAs

• Verify Rfam.cm is current version
• Check that cmpress was run successfully (look for .i1m, .i1p, .i1f, .i1i files)
• For mitochondrial/chloroplast ncRNAs, search organelle sequence separately

tRNAscan-SE Finds Too Many Pseudogenes

• Normal for eukaryotic genomes (especially mammals)
• Filter by score: high-confidence tRNAs typically score > 50 bits
• Use --detail flag to identify pseudogenes in output

Related Skills

• prokaryotic-annotation - Bakta includes ncRNA annotation
• eukaryotic-gene-prediction - Gene prediction excludes ncRNAs
• rna-structure/ncrna-search - Targeted ncRNA homology searches