---
task_categories:
- question-answering
- tabular-classification
- text-generation
language:
- en
tags:
- biology
- proteins
- amino-acids
size_categories:
- 100K<1M
extra_gated_prompt: "Access to this dataset requires a purchase [here](https://buy.stripe.com/6oEbJu5tPci79IQcMX)"
extra_gated_fields:
 Name: text
 Affiliation: text
 Email: text
 I have purchased a license: checkbox
---
# Protein Data Stability - Single Mutation

This repository contains data on the change in protein stability with a single mutation.

There are two datasets:
- [Sample dataset, ~100 datapoints](https://huggingface.co/datasets/Trelis/protein_stability_single_mutation_SAMPLE).
- [Gated dataset, ~250k datapoints](https://huggingface.co/datasets/Trelis/protein_stability_single_mutation).

## Attribution of Data Sources

- **Primary Source**: Tsuboyama, K., Dauparas, J., Chen, J. et al. Mega-scale experimental analysis of protein folding stability in biology and design. Nature 620, 434–444 (2023). [Link to the paper](https://www.nature.com/articles/s41586-023-06328-6)
- **Dataset Link**: [Zenodo Record](https://zenodo.org/record/7992926)

As to where the dataset comes from in this broader work, the relevant dataset (#3) is shown in `dataset_table.jpeg` of this repository's files.

## Sample Protein Stability Data

| Base Protein Sequence                                      | Mutation | ΔΔG_ML            | Classification |
|------------------------------------------------------------|----------|--------------------|----------------|
| DEIHIHFNGVTIEFRGLTEEAKKALEEAIKRGASEEELKELARRLIK            | base     | -0.0675543480388345| neutral       |
| DEIHIHFNGVTIEFRGLTEEAKKALEEAIKRGASEEELKELARRLIK            | D1Q      | -0.0162349479755414| neutral       |
| DEIHIHFNGVTIEFRGLTEEAKKALEEAIKRGASEEELKELARRLIK            | D1E      | -0.1402534094665108| neutral       |
| DEIHIHFNGVTIEFRGLTEEAKKALEEAIKRGASEEELKELARRLIK            | I3K      | -0.6219710162367509| stabilising   |
| DEIHIHFNGVTIEFRGLTEEAKKALEEAIKRGASEEELKELARRLIK            | I3W      |  0.6204920350860421| destabilising |


## Dataset Structure

The dataset focuses on the differential deltaG (mutation minus base) of various protein mutations.

- **Base Protein Sequence** (`aa_seq`): A (sometimes shortened) amino acid sequence.
- **Mutation**: Represented as a combination of amino acid and its position (e.g., F10N indicates changing the 10th amino acid (F) in a sequence for N).
- **delta deltaG** (`ddG_ML`): Derived from a model that makes use of stability measurements measured by two proteases, trypsin and chymotrypsin.
- **Classification**: Classification is done purely on the basis of ddG. The standard deviation of ddG in the dataset is measured and:
-- Rows above 0.5 standard deviations are classified as 'destabilising'
-- Rows below -0.5 standard deviations are classified as 'stabilising'
-- Rows between -0.5 and 0.5 standard deviations are classified as 'neutral'
  
### Understanding ΔG (delta G)

ΔG is the Gibbs free energy change of a process, dictating whether a process is thermodynamically favorable:

- **Negative ΔG**: Indicates the process is energetically favorable. For protein folding, it implies the protein is stable in its folded form.
- **Positive ΔG**: Indicates the process is not energetically favorable. In protein folding, it means the protein requires energy to maintain its folded state.

The **delta delta G** (ΔΔG) represents the deltaG of the mutation compared to the base protein:

- **Positive ΔΔG**: Suggests the mutation decreases protein stability.
- **Negative ΔΔG**: Suggests the mutation enhances protein stability.

### Data Cleanup and Validation:

1. Filtering: The dataset has been curated to only include examples of single mutations.
2. Sequence mutations were extracted from the row names. Base mutations are labelled as 'base'.
3. Consistency Check: Only rows with a consistent 'mutation', aligned with both the base and mutated sequences from the raw data, have been retained.