RNA2HLA

HLA-based quality control of RNA-seq datasets

Synopsis

Tool extracts the HLA types of I and II classes from all the files in the folder containing raw RNA-seq data (paired- or single-end). The alleles are then cross-compared between the RNA-seq samples to identify the common source of the samples based on HLA types (4 digital resolution).

Releases: v1.0 - original tool; v1.1 - script creating heatmap output is added; v1.2 - Python 3.7 port with compatibility and reproducibility fixes

Author

Dr. Irina Chelysheva, 2019-2026 (c)
Senior Lecturer in Human Genomics, Oxford Brookes University
Oxford Vaccine Group, Department of Paediatrics, University of Oxford
Contact

Usage

$ python RNA2HLA.py -f /raw_RNAseq_data_folder [-r /global_name_of_run] [-p <int>] [-3 <int>] [-c <float>] [-g <int>]

Optional parameters:

• -r to be used as a prefix for all output files
• -p number of parallel search threads for bowtie (default: 6)
• -3 trim bases from the low-quality end of each read
• -c confidence level for HLA-typing (default: 0.05)
• -g number of HLA genes to be included for typing (default: 5, may be increased to 6 - adding DQB1)

Dependencies

1. RNA2HLA is a Python 3 script (v1.2, developed and validated with Python 3.7). The Python 2 version is archived in release v1.1.
2. All the dependencies provided within RNA2HLA repository (Python scripts single_end.py and paired_end.py, function scripts in R and Python, HLA class I and II databases) must be downloaded and located in the same folder.
3. Index files must be downloaded and located in subfolder /references.
4. The easiest way to run RNA2HLA is to create a conda environment using the RNA2HLA_env.yml file provided:

$ conda env create -f RNA2HLA_env.yml

And activate it:

$ conda activate RNA2HLA_env

Otherwise:
4a) bowtie (1.x) must be reachable by the command bowtie (validated with versions 1.1.2 and 1.3.1)
4b) R must be installed (validated with R 4.4.3; only base stats functions are used).
4c) Packages: biopython (validated with 1.79), numpy (validated with 1.21.6), pandas (validated with 1.3.5)

Output

The final output - overall comparison matrix in csv format, which cross-compares all RNA-seq samples in the given folder.

Individual outputs in txt format produced for each RNA-seq sample in the folder (classes I and II are written in one file):

1. .bowtielog.txt - file with statistics of HLA mapping;
2. .ambiguity.txt - reports typing ambiguities (if more than one solution for an allele possible based on the expression and HLA databases);
3. .expression.txt - RPKM expression of HLA;
4. .HLAgenotype4digits.txt - 4 digital HLA type.

A heatmap can be created from the overall comparison matrix csv file using the R script heatmap_HLA_identity_comparison.R

What is new in v1.2

• Python 3.7 port. The tool was ported from Python 2.7 and validated side-by-side against the original on real RNA-seq data (6 whole-blood samples from 3 participants x 2 time points, SRP276081, plus a full-depth check): HLA calls, expression values, ambiguity reports and the sample-comparison matrix are reproduced; bowtie mapping logs are byte-identical between versions.
• bowtie >= 1.2 compatibility fix. Previous versions crash with bowtie >= 1.2 because the parser assumed a fixed first line of the bowtie log (bowtie 1.2+ prepends a deprecation warning). The log is now parsed by content, and the tool works with both bowtie 1.1.2 and 1.3.1 with identical results.
• Deterministic tie-breaking. When two allele groups receive exactly the same number of reads, the choice between them was arbitrary in previous versions (dependent on Python 2 dict hash order). Ties are now resolved deterministically (lexicographic order). Note: in the rare event of an exact read-count tie, v1.2 can therefore report a different (but reproducible) allele than v1.0/v1.1.
• Confidence formatting is numerically identical to the Python 2 output (12 significant digits).

Limitation

In the case of studying a particular population with prior knowledge of the low HLA allele diversity, RNA2HLA should not be used as a QC, but only as a convenient study-wide HLA-typing method. One can refer to the Allele Frequency Net Database and discover HLA diversity of particular population through the interactive map. The populations with less than 50 of total known alleles should be considered as of low diversity.

Note: HLA class II second-allele calls are sensitive to sequencing depth; low-confidence calls (p > 0.05) are excluded from the comparison matrix.

Version history

1.0: initial tool
1.1: script creating heatmap output is added
1.2: Python 3.7 port; bowtie >= 1.2 compatibility fix; deterministic tie-breaking

Citations - RNA2HLA

Please, cite the following publication, if you are using RNA2HLA in your research: Irina Chelysheva, Andrew J Pollard, Daniel O'Connor, RNA2HLA: HLA-based quality control of RNA-seq datasets, Briefings in Bioinformatics, 2021

License

MIT