Structure conditioned hallucinated CDR sequences
收藏资源简介:
<em><strong>Datasets accompanying publication https://doi.org/10.1101/2022.06.06.494991</strong></em> (Published in Frontiers in Immunology). Please refer to the final version of the manuscript on Frontiers. <strong>RAbD dataset (Figure 2 in preprint/publication)</strong>: Structure conditioned hallucinated sequences for all 6 CDR loops of 60 antibodies (see publication above for more details) with wildtype seeding (runs_wtseed) and without wildtype seeding (runs_noseed). Full sequences are under runs_<>/<benchamark_name>/<cdr>/results/sequences.fasta. <strong>DeepAb Testset (SI Table 1, SI Figure 4 in preprint/publication)</strong>: Structure conditioned hallucinated sequences for all 6 CDR loops of 20 antibodies selected from the DeepAb test set (see publication above for more details) with wildtype seeding (runs_wtseed) and without wildtype seeding (runs_noseed). Full sequences are under runs_<>/<benchamark_name>/<cdr>/results/sequences.fasta. <strong>Trastuzumab hallucination in various modes described in the manuscript.</strong> 1. <strong>Unrestricted hallucination</strong>: In <strong>"unrestricted.tar"</strong>. Contains hallucination results in the folder "results". Forward folded structures and metrics in "forward_folding", results of virtual screening in "virtual_binding", results of filtering for both folding and virtual screening in "results_filtered_output", and the results from comparison of multiple folding methods (DeepAb and IgFold) in "results_folding_methods". 2. <strong>Motif-restricted hallucination (positions 95, 100A on heavy chain)</strong>: In "<strong>res2pos_95and100A.tar"</strong>. Contains hallucination results in the folder "results". Forward folded structures and metrics in "forward_folding", results of virtual screening in "virtual_binding", results of filtering for both folding and virtual screening in "results_filtered_output", and the results from comparison of multiple folding methods (DeepAb and IgFold) in "results_folding_methods". 3. <strong>Motif-restricted hallucination (positions 99, 100A on heavy chain)</strong>: In "<strong>res2pos_99and100A.tar</strong>". Contains hallucination results in the folder "results". Forward folded structures and metrics in "forward_folding", results of virtual screening in "virtual_binding", results of filtering for both folding and virtual screening in "results_filtered_output", and the results from comparison of multiple folding methods (DeepAb and IgFold) in "results_folding_methods". 4. <strong>Motif-restricted hallucination (positions 95, 99, 100, 100A on heavy chain)</strong>: In "<strong>res4pos.tar</strong>". Contains hallucination results in the folder "results". Forward folded structures and metrics in "forward_folding", results of virtual screening in "virtual_binding", results of filtering for both folding and virtual screening in "results_filtered_output", and the results from comparison of multiple folding methods (DeepAb and IgFold) in "results_folding_methods". 5. <strong>Motif-restricted (positions 99, 100A on heavy chain) and sequence-restricted (restricted to wildtype sequence) hallucination</strong>: In "<strong>res2pos_95and100A_and_sequence.tar</strong>". Contains hallucination results in the folder "results". Forward folded structures and metrics in "forward_folding", results of virtual screening in "virtual_binding", results of filtering for both folding and virtual screening in "results_filtered_output", and the results from comparison of multiple folding methods (DeepAb and IgFold) in "results_folding_methods". For reproducing data, refer to code and methods in the preprint/published version.



