Data and analysis workbooks for Ikonomova et al.-- Experimental Evaluation of AI-Driven Protein Design Risks Using Safe Biological Proxies
May 17, 2025 · View on GitHub
This repository contains the data and workbooks used to create figures contained in the publication "Experimental Evaluation of AI-Driven Protein Design Risks Using Safe Biological Proxies." The files are organized as follows:
.
├── analysis
│ ├── blast_bestmatch.ipynb
│ ├── heatmap.ipynb
│ ├── pdz_analysis.ipynb
│ ├── t7_analysis.ipynb
│ └── ura3_analysis.ipynb
├── CODEMETA.yaml
├── CODEOWNERS
├── data
│ ├── assay
│ │ ├── pdz_all_results_df.csv
│ │ ├── t7_cfe_hierarchical_model.csv
│ │ └── URA3
│ │ ├── 2025-01-03_URA-assay_6-variants
│ │ ├── 2025-01-09_URA-assay_12-variants
│ │ ├── 2025-01-15_URA-assay_16-variants
│ │ ├── 2025-01-16_URA-assay_16-variants
│ │ ├── 2025-01-18_URA-assay_16-variants
│ │ ├── 2025-01-24_URA-assay_16-variants
│ │ ├── 2025-01-25_URA-assay_16-variants
│ │ ├── 2025-01-28_URA-assay_16-variants
│ │ ├── 2025-03-14_URA-assay_16-variants
│ │ ├── 2025-03-18_URA-assay_16-variants
│ │ ├── 2025-03-25_URA-assay_16-variants
│ │ └── 2025-04-22_URA-assay_16-variants
│ ├── blast
│ │ ├── psd95pdz3_blast_results.csv
│ │ ├── t7rnapol_blast_results.csv
│ │ └── ura3_blast_results.csv
│ ├── fasta
│ │ ├── psd95pdz3.fasta
│ │ ├── t7rnapol.fasta
│ │ └── ura3.fasta
│ ├── output
│ │ ├── pdz_blast_flags.csv
│ │ ├── pdz_insilico_merged.csv
│ │ ├── t7_insilico_merged.csv
│ │ ├── t7rnapol_blast_flags.csv
│ │ ├── ura3_blast_flags.csv
│ │ └── ura3_insilico_merged.csv
│ ├── selected_sequences.csv
│ └── ura3_plasmid_list_with_sequence.csv
├── fair-software.md
├── LICENSE.md
├── README.md
└── requirements.txt
/analysiscontains the workbooks;/datacontains all data.requirements.txtcontains all necessary Python modules to run the workbooks.- All Jupyter notebooks are tested on MacOS 15.4, Python 3.12.
Analysis
blast_bestmatch
This workbook generates the csv files containing best matches to synthetic homologs using NCBI BLAST+. Results are saved to data/blast and used in other analysis notebooks.
heatmap
This workbook generates the heatmap from publication Figure 3, using normalized primary assay data from all protein targets.
{protein}_analysis
These workbooks generate the individual plots for the synthetic homolog primary assay results, used to create publication Figures 2-4 and various Supplementary Figures.
seqid_vs_tm
This workbook calculates and plots the relationship between number of sequence substitutions from wildtype in synthetic homologs and calculated TM-score when compared to wildtype. An example for PDZ3 is seen in publication Figure S13.
Data
selected_sequences.csv: contains the in silico metrics for all synthetic homologs.
ura3_plasmid_list_with_sequence.csv: contains the DNA plate locations and identifiers for URA3 synthetic homologs, used to map identifiers to assay results.
assay
This folder contains the data collected in primary assays of synthetic homologs.
blast
This folder contains the output of the blast_bestmatch workbook: best match entries for all synthetic homologs, if found. Blank cells indicate a match was not found.
fasta
This folder contains amino acid or DNA fasta sequences for all synthetic homologs, used by blast_bestmatch to calculate best match entries.
output
This folder contains all data exported by the analysis workbooks, including merged experimental and in silico datasets for synthetic homologs and results of BLAST flagging analysis.
Citation
- Ikonomova, S. P., Wittmann, B. J., Piorino, F., Ross, D. J., Schaffter, S. W., Vasilyeva, O., Horvitz, E., Diggans, J., Strychalski, E. A., Lin-Gibson, S., & Taghon, G. J. (2025 May 16). Experimental evaluation of AI-driven protein design risks using safe biological proxies. bioRxiv. https://doi.org/10.1101/2025.05.15.654077
BibTeX:
@unpublished{Ikonomova2025_TEVV,
author = {Ikonomova, Svetlana P. and Wittmann, Bruce J. and Piorino, Fernanda and Ross, David J. and Schaffter, Samuel W. and Vasilyeva, Olga and Horvitz, Eric and Diggans, James and Strychalski, Elizabeth A. and Lin-Gibson, Sheng and Taghon, Geoffrey J.},
title = {Experimental Evaluation of {AI}-Driven Protein Design Risks Using Safe Biological Proxies},
year = {2025},
month = {May},
day = {16},
note = {bioRxiv preprint},
doi = {10.1101/2025.05.15.654077},
url = {https://doi.org/10.1101/2025.05.15.654077}
}
Contact information
For questions pertaining to these files or the publication in general, please contact:
- geoffrey.taghon@nist.gov (Jupyter notebooks, CSV data)
- bwittmann@microsoft.com (AI sequence generation)