Data for: A sitewise model of natural selection on individual antibodies via a transformer-encoder

This record provides processed B-cell receptor (BCR) data from various studies as described in Matsen IV, F. et al. (2025). A sitewise model of natural selection on individual antibodies via a transformer-encoder. Molecular Biology and Evolution. The software to reproduce the results in that publication can be found here. BCR sequences are clustered into clonal families and germline sequences are inferred. For each clonal family, phylogenetic tree inference and reconstruction of ancestral sequences are performed. The parent and child sequences of each branch of a tree form a "parent-child pair" (PCP), which are used for training or evaluating models. Only heavy chain sequences are considered in this work. PCP data files The dnsm_data.tar.gz contains dataframes of PCP from various datasets and simulations. Processed datasets are found in the DATA_DIR/v3/ directory: file name description wyatt-10x-1p5m_fs-all-NoWinCheck_igh_pcp_2024-10-29_NI_noN_no-naive.csv.gz PCPs of BCR sequences from Jaffe et al. (2022). tang-deepshm-prod-NoWinCheck_igh_pcp_2024-10-29_MASKED_NI_noN_no-naive.csv.gz PCPs of BCR sequences from Tang et al. (2022) and Vergani et al. (2017). rodriguez-airr-seq-race-prod-NoWinCheck_igh_pcp_2024-11-12_MASKED_NI_noN_no-naive.csv.gz PCPs of BCR sequences from Rodriguez et al. (2023). Corresponding per clonal familiy IMGT numbering schemes for each dataset are found in DATA_DIR/v3/anarci/. Simulation datasets are found in the DATA_DIR/simulations/v3/ directory: file name description v3convert_dnsm_jaffe+tang_SIM_rodriguez-6-2-25_NI_no-naive.csv.gz PCPs from simulation of BCR sequences along trees inferred from Rodriguez et al. (2023). v3convert_dnsm_jaffe+tang_SIM_v2tang-2025-6-3_NI_no-naive_CONCAT_dnsm_jaffe+tang_SIM_v1jaffebulk-2025-6-3_NI_no-naive.csv.gz PCPs from simulation of BCR sequences along trees inferred from Jaffe et al. (2022), and Tang et al. (2022) and Vergani et al. (2017). v3convert_dnsm_jaffe+tang_SIM_v2tang-2025-6-3_NI_no-naive_CONCAT_dnsm_jaffe+tang_SIM_v1jaffebulk-2025-6-3_NI_no-naive_downsample_50k.csv.gz 50k subset of PCPs from simulation along trees inferred from Jaffe et al. (2022), and Tang et al. (2022) and Vergani et al. (2017). v3convert_dnsm_jaffe+tang_SIM_v2tang-2025-6-3_NI_no-naive_CONCAT_dnsm_jaffe+tang_SIM_v1jaffebulk-2025-6-3_NI_no-naive_downsample_100k.csv.gz 100k subset of PCPs from simulation along trees inferred from Jaffe et al. (2022), and Tang et al. (2022) and Vergani et al. (2017). v3convert_dnsm_jaffe+tang_SIM_v2tang-2025-6-3_NI_no-naive_CONCAT_dnsm_jaffe+tang_SIM_v1jaffebulk-2025-6-3_NI_no-naive_downsample_250k.csv.gz 250k subset of PCPs from simulation along trees inferred from Jaffe et al. (2022), and Tang et al. (2022) and Vergani et al. (2017). PCP data format column description sample_id sample label, where a sample corresponds to an individual family clonal family label within a sample parent_name label of the parent sequence parent_heavy parent heavy chain sequence child_name label of the child sequence child_heavy child heavy chain sequence branch_length branch length computed in IQ-TREE depth the number of edges that the child sequence is away from the naive sequence in the inferred tree distance sum of branch lengths from the naive sequence to the child sequence v_gene_heavy inferred heavy chain V gene j_gene_heavy inferred heavy chain J gene cdr1_codon_start_heavy position of the first nucleotide of the first codon in heavy chain CDR1 cdr1_codon_end_heavy position of the first nucleotide of the last codon in heavy chain CDR1 cdr2_codon_start_heavy position of the first nucleotide of the first codon in heavy chain CDR2 cdr2_codon_end_heavy position of the first nucleotide of the last codon in heavy chain CDR2 cdr3_codon_start_heavy position of the first nucleotide of the first codon in heavy chain CDR3 cdr3_codon_end_heavy position of the first nucleotide of the last codon in heavy chain CDR3 parent_is_naive True/False whether the parent sequence is the naive sequence of the clonal family child_is_leaf True/False whether the child sequence is a leaf node of the tree rate_scale_heavy Correction factor for the heavy chain evolution rate (simulation files only, but not used in the analysis) rate_scale_light Correction factor for the light chain evolution rate (simulation files only, but not used in the analysis)