# Aptamer Particle-Display Selection Data for ML-Guided Aptamer Refinement (NGAL target)

## Facts

- Provider: Google Research and X Development (Bashir, Yang, Dimon, Ferguson et al.)
- License: CC-BY-4.0 (https://github.com/google-research/google-research/blob/master/README.md)
- Format: CSV (in sampled files)
- Stored at: Google Cloud Storage, gs://gresearch/aptamers_mlpd (us)
- Access: public, readable without an account
- DOI: https://doi.org/10.1038/s41467-021-22555-9
- Category: Drug Discovery
- Source: https://doi.org/10.1038/s41467-021-22555-9
- Page: https://scigantic.com/archive/0377a30d-5656-5ee1-899c-3a481650c3b1
- Keywords: Drug Discovery, CSV, Google Cloud Storage

## About

Sequencing and affinity data from the particle-display (PD) aptamer selection campaign that Bashir et al. used to train models predicting DNA aptamer affinity for human NGAL (lipocalin-2). mlpd_input_data_manuscript.csv holds 40-mer DNA sequences with per-pool read counts from the first selection round at four FACS stringency levels (very, high, medium, low), each split into sorted positive and negative populations, plus the model score that nominated the sequence, its cluster and seed identifiers, the parent sequence and mutation type, and which ordered sequence set and walking model produced it. pd_clustered_input_data_manuscript.csv is the clustered form, where reads were grouped by Levenshtein distance five into connected components and represented by the highest-count cluster representative, the form used for model training. truncation_option_seed_scores_manuscript.csv scores truncated variants, the experiment that yielded an aptamer 70% shorter with 1.5 nM affinity. Measured 2026-11-15: three CSV files, about 105 MB in total; raw reads live separately in NCBI SRA under PRJNA672779. Load with pandas and reproduce the models with the notebooks in the aptamers_mlpd folder. Datasets in the google-research repository are CC BY 4.0; cite Bashir et al. 2021, Nature Communications.

## Load the data

```python
import os
import pandas as pd
df = pd.read_csv(os.path.join(os.environ['SCIGANTIC_MOUNT_PATH'], 'mlpd_input_data_manuscript.csv'), nrows=1000)
df.head()
```

Run in a Scigantic notebook on 2026-11-16 against `mlpd_input_data_manuscript.csv`: CSV/TSV: columns ['sequence', 'library', 'multiple_copy_oligo_in_library', 'round1_high_negative', 'round1_high_positive', 'round1_low_negative', 'round1_low_positive', 'round1_medium_negative', 'round1_medium_positive', 'round1_very_negative', 'round1_very_positive', 'input_model_score'], dtypes ['.

From any machine:

```python
import gcsfs
fs = gcsfs.GCSFileSystem(token='anon')
fs.ls('gresearch/aptamers_mlpd/', detail=False)[:10]
```

Lists the files with anonymous access. Runs on any machine, with no cloud account.

## Data structure

File types in sampled files: CSV

Example file: `aptamers_mlpd/mlpd_input_data_manuscript.csv`

## Open it

- In a hosted notebook with the data mounted: https://scigantic.com/archive/0377a30d-5656-5ee1-899c-3a481650c3b1
- Full record as JSON (schema card included): https://api.scigantic.com/api/archives/0377a30d-5656-5ee1-899c-3a481650c3b1
- Over MCP: `get_schema_card` with id `0377a30d-5656-5ee1-899c-3a481650c3b1` at https://api.scigantic.com/mcp

## Related topics

- [CSV datasets](https://scigantic.com/topics/csv)
- [Drug Discovery datasets](https://scigantic.com/topics/drug-discovery)
