SOURCE DATA: ANTIBIOTIC-TREATED MICROBIAL COMMUNITIES

Paper
Engineering microbial consortia for distributed signal processing
Nature Communications manuscript NCOMMS-26-017824B

Folder / archive
05_Antibiotic_communities

Dataset description
This archive contains the experimental time-course data used to infer amoxicillin and beta-lactamase inhibitor concentrations from two-member microbial communities. The communities contained an amoxicillin-resistant population and an amoxicillin-susceptible population. The resistant population carried one of four beta-lactamase plasmids and constitutively expressed GFP; BFP and optical density provided additional community-level readouts.

The measurements in this archive were generated previously and are reused in the present study. The original broader dataset is supplied in allexps.zip. The eight manuscript-specific datasets in the Data files directory were extracted directly from allexps.csv using the included processing notebook.

Abbreviations and experimental variables
- AMX: amoxicillin
- TAZ: tazobactam
- SUL: sulbactam
- Bla: low-copy plasmid expressing periplasmic beta-lactamase
- BlaM: low-copy plasmid expressing cytoplasmic beta-lactamase
- HSGBla: high-copy plasmid expressing periplasmic beta-lactamase
- HSGBlaM: high-copy plasmid expressing cytoplasmic beta-lactamase
- OD: optical density
- GFP: green fluorescent protein signal
- BFP: blue fluorescent protein signal

Archive contents
1. allexps.zip
   - Contains allexps.csv, the broader original experimental table.
   - allexps.csv contains the condition metadata and time-resolved OD600, GFP, and BFP measurements from the original antibiotic-response experiments.
   - The broader archive contains experiments beyond the eight mixed-community datasets analyzed in the present manuscript.

2. process Helena Ma data allexps.ipynb
   - Extracts manuscript-specific mixed-community datasets from allexps.csv.
   - Filters the original table by plasmid and inhibitor.
   - Writes one condition table and four NumPy arrays for each selected dataset.
   - To rerun the notebook, first extract allexps.csv from allexps.zip into the same working directory.

3. unique_rows.xlsx
   - Reference table containing the distinct mixed-community condition and replicate records present in the broader original dataset.
   - Includes conditions not analyzed in the present manuscript.

4. Data files/
   Contains eight processed manuscript-specific datasets:
   - Bla_SUL
   - Bla_TAZ
   - BlaM_SUL
   - BlaM_TAZ
   - HSGBla_SUL
   - HSGBla_TAZ
   - HSGBlaM_SUL
   - HSGBlaM_TAZ

Dataset size and design
Each of the eight processed datasets contains:
- 300 experimental time courses
- 100 unique AMX-inhibitor concentration combinations
- 3 replicate wells per concentration combination
- 145 measurements per time course
- approximately 24 h of observation, with measurements taken at approximately 10-min intervals

For each dataset, AMX and the corresponding inhibitor were each tested at:
0, 0.5, 1, 2, 4, 8, 16, 32, 64, and 128 micrograms per milliliter.

Across all eight processed datasets, the archive contains 2,400 experimental time courses. These processed files are manuscript-specific extracts of allexps.csv and are included to make the Source Data archive directly usable without requiring readers to reprocess the broader original table.

Processed file organization
Each dataset directory contains five files. For example, HSGBla_TAZ contains:

- HSGBla_TAZ_conditions.csv
- HSGBla_TAZ_od.npy
- HSGBla_TAZ_gfp.npy
- HSGBla_TAZ_bfp.npy
- HSGBla_TAZ_time.npy

The same naming convention is used for all eight datasets.

Condition table
Each *_conditions.csv file contains 300 rows. Row i corresponds exactly to row i of the associated OD, GFP, and BFP arrays.

Columns:
1. strain_background
2. plasmid_name
3. mixed_flag
4. antibiotic_name
5. inhibitor_name
6. [A]: amoxicillin concentration in micrograms per milliliter
7. [I]: inhibitor concentration in micrograms per milliliter
8. Replicate

Time-course arrays
- *_od.npy: shape (300, 145); OD600 time courses
- *_gfp.npy: shape (300, 145); GFP time courses
- *_bfp.npy: shape (300, 145); BFP time courses
- *_time.npy: shape (145,); elapsed time in hours

Within each dataset, row ordering is identical across the condition table and all three measurement arrays. The single time vector applies to every row in that dataset.

Figure mapping
- Main Figure 5b and 5d:
  HSGBla_TAZ

- Main Figure 5c and 5e:
  HSGBla_SUL

- Supplementary Figures 14 and 15:
  Representative antibiotic-response measurements and model-fitting analyses derived from the manuscript-specific processed datasets in this archive.

- Supplementary Figure 16:
  a, Bla_SUL
  b, BlaM_SUL
  c, HSGBla_SUL
  d, HSGBlaM_SUL
  e, Bla_TAZ
  f, BlaM_TAZ
  g, HSGBla_TAZ
  h, HSGBlaM_TAZ

- Supplementary Table 3:
  Crosstalk statistics for all eight plasmid-inhibitor datasets.

Figure 5a is a schematic and does not have associated numerical source data.

Data processing and exclusions
The processed NumPy arrays were extracted from the original allexps.csv table without changing the correspondence between condition records and time-course measurements.

As described in the Methods, specified machine-learning analyses excluded samples with zero input concentration and applied the stated dynamic-range criteria. The complete 10 x 10 concentration matrices, including zero-input conditions used for baseline, dose-response, and crosstalk analyses, are retained in this archive.

The model fits, augmented simulations, latent representations, and predicted concentrations shown in the manuscript are computationally derived from these experimental measurements. They can be regenerated using the archived code, fitted parameters, trained models, and fixed analysis files.

Code and reproducibility
Analysis code is available at:
https://github.com/youlab/multiplexed_sensing

Relevant repository notebooks include:
- figure_5bd_hsgbla_taz.ipynb
- figure_5ce_hsgbla_sul.ipynb
- supplementary_fig_16.ipynb
- process Helena Ma data allexps.ipynb

The analyses were performed using Python 3.12.7.

Contact
Lingchong You
Department of Biomedical Engineering, Duke University
Email: you@duke.edu
