Leveraging T-B reciprocity to enhance B cell epitope prediction
The ability to predict B cell epitopes from antigen sequences is critical for biomedical research and many clinical applications. However, despite substantial efforts over the past 20 years, the performance of even the best B cell epitope prediction software is still modest. Based on the idea of T-B reciprocity, BepiTBR is a B cell epitope prediction model that demonstrates improved performance by incorporating prediction of nearby CD4+ T cell epitopes close to the B cell epitopes.
Researchers interested in more information about BepiTBR and other bioinformatics tools can visit Dr. Tao Wang's lab website.
A free online version of BepiTBR is provided as a part of the Database for Actionable Immunology (dbAI) to facilitate its use by researchers. However, running a local version of BepiTBR is recommended for users with large numbers of samples.
BepiTBR requires a linux x86-64 operating system with basic utilities (tested on RHEL 6, kernel 3.10.0-693 and Ubuntu 18.04, 20.04).
BepiTBR is written in python, raku, and R. It and can be downloaded from github. Simply extract and place the BepiTBR directory in a desired location. Note that some dependecies need to be manually installed. Although actual compile/install times are low, obtaining and installing the dependencies may take approximately 1 hour. ~5 GB of free disk space is also required during installation.
Raku v6.d or later
python 3.6.4+
R 3.6+, glmnet package
conda 4.4.10+
java 1.6+
PERL 5.0+
gcc 5.4.0+
g++ 4.8.5+
tar
download link
follow the instructions in bepipred-1.0.readme to complete installation.
follow the instructions in bp2_env_install_instructions.txt under the install directory to install the bp2 conda environment.
download site: https://github.com/brsaran/LBEEP
follow the instructions in README.md to install LBEEP 1.0.
download site: http://www.cbs.dtu.dk/cgi-bin/nph-sw_request?netMHCIIpan
follow the instructions in netMHCIIpan-3.2.readme to complete the installation.
download site: https://github.com/GfellerLab/MixMHC2pred
follow the instructions in README.md to install.
This tutorial will guide the user in running BepiTBR in several different modes. The generated results can be compared to the provided results in examples/example_output
Analyzes specific epitopes within proteins
Requires epitope sequences and their corresponding full antigen protein sequences
Example command using provided example files:
(replace paths with appropriate paths according to user's particular installation)
raku BepiTBR.raku \
--motif0_file=examples/test_data_BepiTBR/Ind-positive.txt \
--full0_file=examples/test_data_BepiTBR/peptide_with_full_length.txt \
--bepipred2=bp2/bin/activate \
--bepipred1=/home/exampleUser/bp1/bepipred-1.0/bepipred \
--LBEEP=/home/exampleUser/LBEEP/ \
--MixMHC2pred=/home/exampleUser/MixMHC2pred/MixMHC2pred_unix \
--netMHCIIpan=/home/exampleUser/netMHCIIpan-3.2/netMHCIIpan \
--dir=/home/exampleUser/BepiTBR/examples/test_output_BepiTBR \
--thread=4
--motif0_file: candidate epitopes file
--full0_file: full protein sequences file
--bepipred2: the conda environment activation file for BepiPred 2.0 (<bp2_env_directory>/bin/activate)
--bepipred1: path to the BepiPred 1.0 executable
--LBEEP: path to the LBEEP directory
--MixMHC2pred: path to the MixMHC2pred executable
--netMHCIIpan: path to the NetMHCIIpan 3.2 executable
--dir: output directory
--thread: number of CPU thread to use; system dependent
The main output is the predictions.txt file in the output directory. Each line corresponds to an epitope and consists of tab-seperated data with the format:
epitope name, base bepipred1.0, base bepipred2.0, base LBEEP, enhanced bepipred1.0, enhanced bepipred2.0, enhanced LBEEP, ensemble, epitope sequence. For example:
base_bepipred1.0 base_bepipred2.0 base_LBEEP enhanced_bepipred1.0 enhanced_bepipred2.0 enhanced_LBEEP ensemble epitope
Negative_IEDB_ID_946463 1.927 0.6338 0.67 -0.5321 -0.3189 -0.4096 0.8883 KKLIPNPNKIRKPPK
Negative_IEDB_ID_947078 1.013 0.6201 0.51 -0.5706 -0.4672 -0.6792 0.5832 VTRLRYRSVREVWQS
...
Here, base indicates original prediction scores using existing B cell epitope prediction software, enhanced indicates predictions made by considering both the base models and T cell epitope predictions, for each of the three base models, and ensemble is the final BepiTBR prediction score that is the aggregate of the three enhanced models.
The B cell and T cell (MHC II) predictions used to calculate the final scores are removed by default in order to save space; they can be kept using --keep=True.
The example code should take ~4 minutes to complete. Performance can be increased by increasing --thread if hardware resources allows. Peak memory usage is approximately 3 GB/thread in this mode.
Identify potential B cell epitopes from a full antigen protein sequence
Requires the antigen protein sequence to be input directly
Example command:
(replace paths with appropriate paths according to user's particular installation)
raku BepiTBR_full.raku \
--full0=$(cat examples/test_data_BepiTBR_full/example_full.txt) \
--length=15 \
--bepipred2=bp2/bin/activate \
--bepipred1=/home/exampleUser/bp1/bepipred-1.0/bepipred \
--LBEEP=/home/exampleUser/LBEEP/ \
--MixMHC2pred=/home/exampleUser/MixMHC2pred/MixMHC2pred_unix \
--netMHCIIpan=NA \
--dir=/home/exampleUser/BepiTBR/example/test_output_BepiTBR_full \
--thread=8
--full0: protein sequence
--length: length of the B cell epitopes to scan in a moving window. Recommended length: 15
--bepipred2: the conda environment activation file for BepiPred 2.0 (<bp2_env_directory>/bin/activate)
--bepipred1: path to the BepiPred 1.0 executable
--LBEEP: path to the LBEEP directory
--MixMHC2pred: path to the MixMHC2pred executable
--netMHCIIpan: recommended to set to NA to greatly improve run speed
--dir: output directory
--thread: number of CPU thread to use; system dependent
The main output is the predictions.txt file in the output directory. Each line corresponds to the start location of a potential B cell epitope on the protein and consists of tab-seperated data with the format:
epitope name, base bepipred1.0, base bepipred2.0, base LBEEP, enhanced bepipred1.0, enhanced bepipred2.0, enhanced LBEEP, ensemble, epitope sequence. For example:
base_bepipred1.0 base_bepipred2.0 base_LBEEP enhanced_bepipred1.0 enhanced_bepipred2.0 enhanced_LBEEP ensemble epitope
job_pos=0 1.855 0.6331 0.47 -0.4154 -0.5009 -0.8526 0.5792 MTENSTSAPAAKPKR
job_pos=1 1.855 0.6418 0.48 -0.4242 -0.4819 -0.8472 0.5914 TENSTSAPAAKPKRA
job_pos=2 1.855 0.6418 0.48 -0.433 -0.4867 -0.8582 0.5743 ENSTSAPAAKPKRAK
...
same as epitope mode, base indicates original prediction scores using existing B cell epitope prediction software, enhanced indicates predictions made by considering both the base models and T cell epitope predictions, for each of the three base models, and ensemble is the final BepiTBR prediction score that is the aggregate of the three enhanced models.
Same as epitope mode, the B cell and T cell (MHC II) predictions used to calculate the final scores are compressed in order to save space; they are kept as Bepi.tar.gz and Tepi.tar.gz, respectively.
The example code should take 1-2 minutes to complete. Unlike epitope mode, memory use does not scale with --thread, thus using the maximum number of CPU cores available is recommended.
Identify potential B cell epitopes from a .fasta file containing multiple antigen protein sequences
Requires protein sequence file in fasta format
Example command:
(replace paths with appropriate paths according to user's particular installation)
raku BepiTBR_fasta.raku \
--fasta0=examples/test_data_BepiTBR_fasta/peptide_with_full_length.txt \
--length=15 \
--bepipred2=bp2/bin/activate \
--bepipred1=/home/exampleUser/bp1/bepipred-1.0/bepipred \
--LBEEP=/home/exampleUser/LBEEP/ \
--MixMHC2pred=/home/exampleUser/MixMHC2pred/MixMHC2pred_unix \
--netMHCIIpan=NA \
--dir=/home/exampleUser/BepiTBR/examples/test_output_BepiTBR_fasta \
--thread=4
--fasta0: protein sequence file in fasta format
--length: length of the B cell epitopes to scan in a moving window. Recommended length: 15
--bepipred2: the conda environment activation file for BepiPred 2.0 (<bp2_env_directory>/bin/activate)
--bepipred1: path to the BepiPred 1.0 executable
--LBEEP: path to the LBEEP directory
--MixMHC2pred: path to the MixMHC2pred executable
--netMHCIIpan: recommended to set to NA to greatly improve run speed
--dir: output directory; results for each antigen sequence will be placed in a sub-directory
--thread: number of CPU thread to use; system dependent
For each protein in the fasta file labelled with a header line, a sub-directory for that protein will be created in the output directory. For example, for the input in examples/test_data_BepiTBR_fasta, two sub-directories, Positive_IEDB_ID_114414 and Positive_IEDB_ID_123439 will be created in the output directory. In each sub-directory, the main output is the predictions.txt file. Same as for full protein mode, each line corresponds to the start location of a potential epitope on the protein and consists of tab-seperated data with the format:
epitope name, base bepipred1.0, base bepipred2.0, base LBEEP, enhanced bepipred1.0, enhanced bepipred2.0, enhanced LBEEP, ensemble, epitope sequence. For Positive_IEDB_ID_114414, the example output is as follows:
base_bepipred1.0 base_bepipred2.0 base_LBEEP enhanced_bepipred1.0 enhanced_bepipred2.0 enhanced_LBEEP ensemble epitope
job_pos=0 0.656 0.3888 0.43 -0.8591 -1.2612 -1.2655 -0.6383 MASSSSVLLVVALFA
job_pos=1 0.656 0.3888 0.32 -0.844 -1.2423 -1.583 -0.7871 ASSSSVLLVVALFAV
job_pos=2 0.288 0.3888 0.29 -0.9434 -1.235 -1.6714 -0.9076 SSSSVLLVVALFAVF
...
and for Positive_IEDB_ID_123439:
base_bepipred1.0 base_bepipred2.0 base_LBEEP enhanced_bepipred1.0 enhanced_bepipred2.0 enhanced_LBEEP ensemble epitope
job_pos=0 -1.393 0.3664 0.28 -1.891 -1.1241 -1.5599 -1.4901 MRLLQCVLLCVSLSL
job_pos=1 -0.917 0.3923 0.3 -1.6216 -1.0555 -1.4938 -1.1855 RLLQCVLLCVSLSLV
job_pos=2 -0.719 0.4382 0.32 -1.5136 -0.9351 -1.4293 -0.9637 LLQCVLLCVSLSLVL
...
same as the other two modes, base indicates original prediction scores using existing B cell epitope prediction software, enhanced indicates predictions made by considering both the base models and T cell epitope predictions, for each of the three base models, and ensemble is the final BepiTBR prediction score that is the aggregate of the three enhanced models.
Also the same as the other two modes, the B cell and T cell (MHC II) predictions used to calculate the final scores are compressed in order to save space; they are kept as Bepi.tar.gz and Tepi.tar.gz, respectively.
A combined output is also provided in the main output directory.
The example code should take ~7 minutes to complete. Using the maximum number of CPU cores available / 4 is recommended.
The training/validation data used for BepiTBR were downloaded from IEDB and then filtered according to a number of criteria. It is provided on the BepiTBR github page.
The epitope predictions for select SARS-CoV-2 sequences are available on the BepiTBR github page (reference sequence predictions, B epitopes, T epitopes). For the reference sequence predictions file, "label" refers to the protein/ORF name, "epitope_start_position" the protein position of the first amino acid in potential epitopes, and the other columns are the same as those in bepiTBR outputs. The suggested cutoff for epitopes are 0 for general epitopes, and 0.5 for highly likely epitopes (referring to the "ensemble" column). The original viral sequences were downladed from NGDC and GISAID. The original sequences are not provided here as per GISAID's terms of use. Due to large file sizes, these two files were uploaded using git lfs. To download the files, use git clone or git pull once the git lfs extension is installed.
If you use BepiTBR in your research, please cite our paper.
Zhu J, Gouru A, Wu F, Berzofsky JA. Xie Y, and Wang T. (2022). BepiTBR: T-B reciprocity enhances B cell epitope prediction. iScience. 25. 103764. 10.1016/j.isci.2022.103764.