XBCR-net (Cross B-Cell Receptor network) for antibody-antigen binding prediction

[](https://www.nature.com/articles/s41422-022-00727-6) [](https://colab.research.google.com/github/jianqingzheng/XBCR-net/blob/main/XBCR_net.ipynb)

Code for Cell Research paper Deep learning-based rapid generation of broadly reactive antibodies against SARS-CoV-2 and its Omicron variant

This repo provides an implementation of the training and inference pipeline for XBCR-net based on TensorFlow and Keras. The original implementation of its backbone network ACNN could be found in ACNN repo.

Contents

• 0. Brief Introduction
• 1. Installation
• 2. Usage
  • 2.1. Training (optional)
  • 2.2a. Inference by entering data
  • 2.2b. Batch Inference
• 3. Demo and Tutorial
• 4. Citing this work

0. Brief Intro

• a. The features of the amino acid sequences of VH, VL and RBD sequences were extracted, localized and max-pooled to be concatenated together as input to the fully connected layers. The active features in the latent space were then processed by Multi-Layer Perceptron to predict the binding probability of antibody to multiple antigens. The impact score of VH, VL and RBD is calculated on the local histogram impact score map, representing how much weight is given to the specified amino acids on VH, VL (y axis) and RBD (x axis). Prediction results are evaluated by Precision-Recall curves of ACNN (Violet), Transformer (Gray), FCN (Red) and CNN (Blue).
• b. The HCDR3 sequences of the predicted SARS-CoV-2 and Omicron variant binders are clustered by using an 80% sequence similarity. Cluster size represents the number of BCR sequences in the cluster. For each expanded cluster, the HCDR3 sequences are visualized as a sequence logo plot, where y-axis represents the frequency of the individual amino acid at the corresponding position in x-axis. The frequency of the dominating VH gene is listed above the logo.
• c. Circos plot showing the frequency of antibodies encoded by the specified V region to J region pairing of the pan-SARS2 sequences.
• d. The diversity of the four groups of BCR repertoire is analyzed, which is linked to the sample number of each group.


• e. Binding of the predicted cross-reactive antibodies to RBD of SARS-CoV-2 and Omicron variants (left panel) was examined by ELISA. Representative OD reading is plotted as heatmap ranging from 0.05 to 5.0, and OD of 0.1 is used as cut-off value (n = 3 per group).
• f. The SARS-CoV-2 Omicron variant (BA.1) pseudovirus neutralization curves of XBN-1, XBN-6 and XBN-11 mAbs were generated from luciferase readings at 8 dilutions (n = 3).
• g. HCDR3 sequences of the XBN-1 and XBN-11 are aligned with the most convergent anti-SARS-CoV-2 antibodies from the published studies.
• h. The HCDR3 sequence frequency of the dominant cluster (encoded by IGHV3-30 and IGKV1-13) of the pan-SARS group is shown.

1. Installation

Clone code from Github repo: https://github.com/jianqingzheng/XBCR-net.git

shell git clone https://github.com/jianqingzheng/XBCR-net.git cd XBCR-net/

install packages

shell pip install tensorflow==2.4.1 pip install numpy==1.19.5 pip install pandas==1.1.0

Other versions of the packages could also be applicable

2. Usage

* Setup [$DOWNLOAD_DIR]/XBCR-net/ ├── data/[$data_name]/ | ├── exper/ | | | # experimental dataset for training (.xlsx|.csv files) | | └── example-experimental_data.xlsx | ├── nonexp/ | | | # negative samples for training (.xlsx|.csv files) | | └── example-negative_data.xlsx | └── test/ | ├── ab_to_pred/ | | | # the antibody data for inference | | └── example-antibody_to_predict.xlsx | ├── ag_to_pred/ | | | # the antigen data for inference | | └── example-antigen_to_predict.xlsx | └── results/ | | # the files to print the inference results | └── results_rbd_[$model_name]-[$model_num].xlsx └── models/[$data_name]/ └── [$data_name]-[$model_name]/ | # the files of model parameters (.tf.index and .tf.data-000000-of-00001 files) ├── model_rbd_[$model_num].tf.index └── model_rbd_[$model_num].tf.data-000000-of-00001

Default data can be also downloaded from Data_S1 (unnecessary in usage)

2.1. Training (optional)

1. Upload the experimental data in XBCR-net/data/$data_name/exper/ and the non-experimental data in XBCR-net/data/$data_name/nonexp/
2. Run python main_train.py --model_name XBCR_net --data_name $data_name --model_num $model_num --max_epochs max_epochs --include_light [1/0]
3. Check the saved model in XBCR-net/models/$data_name/$data_name-XBCR_net/

* Example for training (default): 1. Check the experimental data in XBCR-net/data/$data_name/exper/ and the non-experimental data in XBCR-net/data/$data_name/nonexp/ 2. Run shell python main_train.py --model_name XBCR_net --data_name binding --model_num 0 --max_epochs 100 --include_light 1 3. Check the saved model in XBCR-net/models/$data_name/$data_name-XBCR_net/

2.2a. Inference by entering data

* Example for a single data point:

```shell HEAVY='VQLVESGGGLVQPGGSLRLSCAASGFTFSSYDMHWVRQTTGKGLEWVSTIGTAGDTYYPDSVKGRFTISREDAKNSLYLQMNSLRAGDTAVYYCARGDSSGYYYYFDYWGQGTLLTVSS' LIGHT='DIEMTQSPSSLSAAVGDRVTITCRASQSIGSYLNWYQQKPGKAPKLLIYAASSLQSGVPSRFSGSGSGTDFTLTISSLQPEDFAIYYCQQSYVSPTYTFGPGTKVDIK' ANTIG='RVQPTESIVRFPNITNLCPFGEVFNATRFASVYAWNRKRISNCVADYSVLYNSASFSTFKCYGVSPTKLNDLCFTNVYADSFVIRGDEVRQIAPGQTGKIADYNYKLPDDFTGCVIAWNSNNLDSKVGGNYNYLYRLFRKSNLKPFERDISTEIYQAGSTPCNGVEGFNCYFPLQSYGFQPTNGVGYQPYRVVVLSFELLHAPATVCGPKKSTNLVKNKCVNF'

python predbcr.py --heavy $HEAVY --light $LIGHT --antig $ANTIG --modelname XBCRnet --dataname binding --model_num 0 ```

Leave LIGHT="" or LIGHT="_" to exclude the input of light chain.

* Example for multiple data points (split by ','):

```shell HEAVY='VQLVESGGGLVQPGGSLRLSCAASGFTFSSYDMHWVRQTTGKGLEWVSTIGTAGDTYYPDSVKGRFTISREDAKNSLYLQMNSLRAGDTAVYYCARGDSSGYYYYFDYWGQGTLLTVSS, EVQLVESGGGLVQPGGSLRLSCAASGFTFNNYWMSWVRQAPGKGLEWVANINQDGSEKYYVDSVMGRFAISRDNAKNSLYLQMNSLRAEDTAVYYCARDQGYGDYFEYNWFDPWGQGTLVTVSS' LIGHT='DIEMTQSPSSLSAAVGDRVTITCRASQSIGSYLNWYQQKPGKAPKLLIYAASSLQSGVPSRFSGSGSGTDFTLTISSLQPEDFAIYYCQQSYVSPTYTFGPGTKVDIK, DIQLTQSPSFLSASVGDRVTITCRASQGIYSYLAWYQQKPGKAPKLLIYAASTLQSGVPSRFSGSGSGTEFTLTISSLQPEDFATYYCQQLNSYPITFGQGTRLEIK' ANTIG='RVQPTESIVRFPNITNLCPFGEVFNATRFASVYAWNRKRISNCVADYSVLYNSASFSTFKCYGVSPTKLNDLCFTNVYADSFVIRGDEVRQIAPGQTGKIADYNYKLPDDFTGCVIAWNSNNLDSKVGGNYNYLYRLFRKSNLKPFERDISTEIYQAGSTPCNGVEGFNCYFPLQSYGFQPTNGVGYQPYRVVVLSFELLHAPATVCGPKKSTNLVKNKCVNF, RVQPTESIVRFPNITNLCPFGEVFNATRFASVYAWNRKRISNCVADYSVLYNSASFSTFKCYGVSPTKLNDLCFTNVYADSFVIRGDEVRQIAPGQTGKIADYNYKLPDDFTGCVIAWNSNNLDSKVGGNYNYLYRLFRKSNLKPFERDISTEIYQAGSTPCNGVEGFNCYFPLQSYGFQPTNGVGYQPYRVVVLSFELLHAPATVCGPKKSTNLVKNKCVNF'

python predbcr.py --heavy $HEAVY --light $LIGHT --antig $ANTIG --modelname XBCRnet --dataname binding --model_num 0 ```

Set LIGHT="XXX, ,XXX" or LIGHT="XXX,_,XXX" to selectively exclude the input of light chains.\ Spaces (' ' or '_') and carriage returns ('\n') are not recognized as a part of sequence.

2.2b. Batch Inference

1. Upload the antibody file in XBCR-net/data/$data_name/ab_to_pred/ and the antibody file in XBCR-net/data/$data_name/ag_to_pred/
2. Run python main_infer.py --model_name XBCR_net --data_name $data_name --model_num $model_num --include_light [1/0]
3. Download the result Excel file from XBCR-net/data/binding/test/results/*

* Example for inference (default): 1. Check the antibody file in XBCR-net/data/$data_name/ab_to_pred/ and the antibody file in XBCR-net/data/$data_name/ag_to_pred/ 2. Run shell python main_infer.py --model_name XBCR_net --data_name binding --model_num 0 --include_light 1 3. Download the result Excel file from XBCR-net/data/binding/test/results/results_rbd_XBCR_net-0.xlsx

3. Demo and Tutorial

A demo can be found in the provided notebook.

Alternatively, it can be easily run via .

Additionally, a detailed tutorial can be found in the provided (powered by Zread.ai).

4. Citing this work

Any publication that discloses findings arising from using this source code or the network model should cite: - Hantao Lou, Jianqing Zheng, Xiaohang Leo Fang, Zhu Liang, Meihan Zhang, Yu Chen, Chunmei Wang, Xuetao Cao, "Deep learning-based rapid generation of broadly reactive antibodies against SARS-CoV-2 and its Omicron variant." Cell Research 33.1 (2023): 80-82. bibtex @article{lou2022deep, title={Deep learning-based rapid generation of broadly reactive antibodies against SARS-CoV-2 and its Omicron variant}, author={Lou, Hantao and Zheng, Jianqing and Fang, Xiaohang Leo and Liang, Zhu and Zhang, Meihan and Chen, Yu and Wang, Chunmei and Cao, Xuetao}, journal={Cell Research}, pages={1--3}, year={2022}, publisher={Nature Publishing Group}, doi={10.1038/s41422-022-00727-6}, } and, if applicable, the ACNN paper: - Xiao-Yun Zhou, Jian-Qing Zheng, Peichao Li, and Guang-Zhong Yang, "ACNN: a full resolution dcnn for medical image segmentation." 2020 IEEE International Conference on Robotics and Automation (ICRA). IEEE, 2020. bibtex @inproceedings{zhou2020acnn, title={Acnn: a full resolution dcnn for medical image segmentation}, author={Zhou, Xiao-Yun and Zheng, Jian-Qing and Li, Peichao and Yang, Guang-Zhong}, booktitle={2020 IEEE International Conference on Robotics and Automation (ICRA)}, pages={8455--8461}, year={2020}, organization={IEEE}, doi={10.1109/ICRA40945.2020.9197328}, }