orfhunter

name: bioconductor-orfhunterdescription: The ORFhunteR package is a R and C++ library for an automatic determination and annotation of open reading frames (ORF) in a large set of RNA molecules. It efficiently implements the machine learning model based on vectorization of nucleotide sequences and the random forest classification algorithm. The ORFhunteR package consists of a set of functions written in the R language in conjunction with C++. The efficiency of the package was confirmed by the examples of the analysis of RNA molecules frwhen_to_use: Use when: The ORFhunteR package is a R and C++ library for an automatic determination and annotation of open reading frames (ORF); The ORFhunteR package: User’s manual. Not for: Requires R ≥ 4.0.0 and Bioconductor ≥ 3.16user-invocable: false

ORFhunteR

Workflows

Predict And Annotate Orfs

library(ORFhunteR)

trans <- system.file("extdata", "Set.trans_sequences.fasta", package = "ORFhunteR")
model <- "http://www.sstcenter.com/download/ORFhunteR/classRFmodel_1.rds"

# Predict ORFs
ORFs <- predictORF(tr = trans, model = model, prThr = 0)

# Write coordinates to a temporary file for downstream functions
orfs_path <- tempfile(fileext = ".txt")
write.table(ORFs, file = orfs_path, sep = "\t", row.names = FALSE, quote = FALSE)

# Extract sequences
seq_orfs <- getSeqORFs(orfs = orfs_path, tr = trans, genome = "BSgenome.Hsapiens.UCSC.hg38")

# Find PTCs
gtf_path <- system.file("extdata", "Set.trans_sequences.gtf", package = "ORFhunteR")
ptcs <- findPTCs(orfs = orfs_path, gtf = gtf_path)

# Translate ORFs
seq_orf_path <- system.file("extdata", "Set.trans_ORFs.sequences.fasta", package = "ORFhunteR")
prot_seqs <- translateORFs(seqORFs = seq_orf_path)

# Annotate ORFs
prts_path <- system.file("extdata", "Set.trans_proteins.sequences.fasta", package = "ORFhunteR")
anno_orfs <- annotateORFs(orfs = orfs_path, tr = trans, gtf = gtf_path, prts = prts_path)

Input: FASTA file of transcript sequences and a pre-trained Random Forest model.
Output: Predicted ORFs, extracted sequences, PTC status, translated proteins, and a comprehensive annotation table.

Standard Workflow

library(ORFhunteR)

# Load pseudo-ORFs (lncRNAs) and true ORFs (mRNAs)
fileORFLncRNAs <- "http://www.sstcenter.com/download/ORFhunteR/NCBI_RefSeq_release_109_GRCh38.p12_ORF_candidates_sequences_lncRNAs.fasta.gz"
ORFLncRNAs <- loadTrExper(tr = fileORFLncRNAs)

fileORFmRNAs <- "http://www.sstcenter.com/download/ORFhunteR/NCBI_RefSeq_release_109_GRCh38.p12_ORFs_true_sequences_mRNAs.fasta.gz"
ORFmRNAs <- loadTrExper(tr = fileORFmRNAs)

# Train the Random Forest classifier
clt <- classifyORFsCandidates(
  ORFLncRNAs = ORFLncRNAs[1:10],
  ORFmRNAs = ORFmRNAs[1:10],
  pLearn = 0.75,
  nTrees = 10
)

Input: Lists of pseudo-ORFs and true coding ORFs.
Output: A trained randomForest classifier object.

Orf Prediction And Annotation

library(ORFhunteR)

trans <- system.file("extdata", "Set.trans_sequences.fasta", package = "ORFhunteR")
model <- "http://www.sstcenter.com/download/ORFhunteR/classRFmodel_1.rds"

# 1. Predict true ORFs
ORFs <- predictORF(tr = trans, model = model, prThr = 0)

# 2. Extract nucleotide sequences of predicted ORFs
orfs_path <- system.file("extdata", "Set.trans_ORFs.coordinates.txt", package = "ORFhunteR")
seq_orfs <- getSeqORFs(orfs = orfs_path, tr = trans, genome = "BSgenome.Hsapiens.UCSC.hg38")

# 3. Detect premature termination codons (PTCs)
gtf_path <- system.file("extdata", "Set.trans_sequences.gtf", package = "ORFhunteR")
ptcs <- findPTCs(orfs = orfs_path, gtf = gtf_path)

# 4. Translate predicted ORFs into amino acid sequences
seq_orf_path <- system.file("extdata", "Set.trans_ORFs.sequences.fasta", package = "ORFhunteR")
prot_seqs <- translateORFs(seqORFs = seq_orf_path)

# 5. Generate a comprehensive annotation table
prts_path <- system.file("extdata", "Set.trans_proteins.sequences.fasta", package = "ORFhunteR")
anno_orfs <- annotateORFs(orfs = orfs_path, tr = trans, gtf = gtf_path, prts = prts_path)

Input: Transcript sequences, coordinates of ORFs, GTF file, and protein sequences.
Output: Annotated ORF table containing molecular weight, isoelectric point, and potential protein interaction index.

When to Use

• Automatically identifying and annotating open reading frames (ORFs) in large sets of RNA molecules.
• Vectorizing sequence features of ORFs (mono-, di-, trinucleotide frequencies, Bao model entropy, etc.) using vectorizeORFs.
• Training custom Random Forest classifiers to distinguish true ORFs from pseudo-ORFs using classifyORFsCandidates.
• Translating nucleotide sequences of identified ORFs into amino acid sequences using translateORFs.

When NOT to Use

• For general gene annotation or transcript assembly from RNA-seq reads; use StringTie or Scallop instead.
• For predicting protein 3D structures or detailed functional domains; use specialized tools like InterProScan or AlphaFold.

Data Requirements

• Transcript sequences in fasta or fa format (or gtf/gff format with a corresponding BSgenome package like BSgenome.Hsapiens.UCSC.hg38).
• Coordinates of ORFs in a tab-delimited TXT file (as generated by predictORF).
• Transcript structure in gtf or gff format for PTC detection.

Key Parameters

• tr: Character string giving the path to the file with transcripts of interest (FASTA, FA, GTF, or GFF).
• model: Character string giving the path or connection to the pre-trained Random Forest model RDS file.
• prThr (0): Probability threshold for the winning class of ORFs.
• codStart ("ATG"): Start codon to scan for in findORFs.
• pLearn (0.75): Fraction of ORFs used for the training set in classifyORFsCandidates.
• nTrees (500): Number of trees to grow in the Random Forest model.
• genome ("BSgenome.Hsapiens.UCSC.hg38"): Name of the pre-installed BSgenome data package.

Best Practices

• Use loadTrExper to safely load and parse transcript sequences from FASTA or GTF files.
• Filter out low-probability ORF candidates by setting a strict probability threshold prThr (e.g., 0.9) in predictORF to reduce false positives.
• Ensure that the transcript IDs in the coordinates file (orfs), transcript sequences (tr), and GTF file (gtf) match exactly when running annotateORFs or findPTCs.

Common Pitfalls

• Missing BSgenome Package: Running getSeqORFs or predictORF with GTF/GFF inputs fails if the specified genome package is not installed. Fix: Install the required BSgenome package (e.g., BSgenome.Hsapiens.UCSC.hg38) before running.
• Mismatched Transcript IDs: Downstream annotation functions fail to find matching transcripts. Fix: Verify that the transcript IDs in the FASTA, GTF, and coordinates files are identical.
• Low Random Forest Accuracy: Training a model with too few trees or too small a dataset. Fix: Use a larger training set (e.g., N = 1000) and set nTrees to at least 500 in classifyORFsCandidates.

Alternatives

• systemPipeR: For general workflow management including ORF prediction.
• Biostrings: For basic sequence manipulation and translation without machine learning-based ORF classification.

Citations

• Vasily V. Grinev, Mikalai M. Yatskou, Victor V. Skakun, Maryna K. Chepeleva, Petr V. Nazarov (2021). ORFhunteR: an accurate approach for the automatic identification and annotation of open reading frames in human mRNA molecules. bioRxiv.

References

• Homepage: https://bioconductor.org/packages/orfhunter
• Vignette: https://bioconductor.org/packages/release/bioc/vignettes/orfhunter/inst/doc/ORFhunteR.html