README.md

---
configs:
  - config_name: secstr
    data_files:
      - split: train
        path: data/secstr/train.parquet
      - split: validation
        path: data/secstr/validation.parquet
      - split: test
        path: data/secstr/test.parquet
  - config_name: secstr8
    data_files:
      - split: train
        path: data/secstr8/train.parquet
      - split: validation
        path: data/secstr8/validation.parquet
      - split: test
        path: data/secstr8/test.parquet
  - config_name: disorder
    data_files:
      - split: train
        path: data/disorder/train.parquet
      - split: validation
        path: data/disorder/validation.parquet
      - split: test
        path: data/disorder/test.parquet
  - config_name: rsa
    data_files:
      - split: train
        path: data/rsa/train.parquet
      - split: validation
        path: data/rsa/validation.parquet
      - split: test
        path: data/rsa/test.parquet
  - config_name: meltome
    data_files:
      - split: train
        path: data/meltome/train.parquet
      - split: validation
        path: data/meltome/validation.parquet
      - split: test
        path: data/meltome/test.parquet
  - config_name: subloc
    data_files:
      - split: train
        path: data/subloc/train.parquet
      - split: validation
        path: data/subloc/validation.parquet
      - split: test
        path: data/subloc/test.parquet
  - config_name: gpsite_dna
    data_files:
      - split: train
        path: data/gpsite_dna/train.parquet
      - split: validation
        path: data/gpsite_dna/validation.parquet
      - split: test
        path: data/gpsite_dna/test.parquet
  - config_name: gpsite_rna
    data_files:
      - split: train
        path: data/gpsite_rna/train.parquet
      - split: validation
        path: data/gpsite_rna/validation.parquet
      - split: test
        path: data/gpsite_rna/test.parquet
  - config_name: gpsite_pep
    data_files:
      - split: train
        path: data/gpsite_pep/train.parquet
      - split: validation
        path: data/gpsite_pep/validation.parquet
      - split: test
        path: data/gpsite_pep/test.parquet
  - config_name: gpsite_pro
    data_files:
      - split: train
        path: data/gpsite_pro/train.parquet
      - split: validation
        path: data/gpsite_pro/validation.parquet
      - split: test
        path: data/gpsite_pro/test.parquet
  - config_name: gpsite_atp
    data_files:
      - split: train
        path: data/gpsite_atp/train.parquet
      - split: validation
        path: data/gpsite_atp/validation.parquet
      - split: test
        path: data/gpsite_atp/test.parquet
  - config_name: gpsite_hem
    data_files:
      - split: train
        path: data/gpsite_hem/train.parquet
      - split: validation
        path: data/gpsite_hem/validation.parquet
      - split: test
        path: data/gpsite_hem/test.parquet
  - config_name: gpsite_zn
    data_files:
      - split: train
        path: data/gpsite_zn/train.parquet
      - split: validation
        path: data/gpsite_zn/validation.parquet
      - split: test
        path: data/gpsite_zn/test.parquet
  - config_name: gpsite_ca
    data_files:
      - split: train
        path: data/gpsite_ca/train.parquet
      - split: validation
        path: data/gpsite_ca/validation.parquet
      - split: test
        path: data/gpsite_ca/test.parquet
  - config_name: gpsite_mg
    data_files:
      - split: train
        path: data/gpsite_mg/train.parquet
      - split: validation
        path: data/gpsite_mg/validation.parquet
      - split: test
        path: data/gpsite_mg/test.parquet
  - config_name: gpsite_mn
    data_files:
      - split: train
        path: data/gpsite_mn/train.parquet
      - split: validation
        path: data/gpsite_mn/validation.parquet
      - split: test
        path: data/gpsite_mn/test.parquet
license: mit
tags:
  - protein
  - biology
  - benchmark
  - data-leakage
  - protein-language-model
  - multi-task
---

# Moomboh/mutafitup-datasets

Leakage-free protein function prediction benchmarks for multi-task
protein language model (pLM) fine-tuning.

This dataset contains **16 prediction tasks** drawn from
three independent data families, resplit to eliminate cross-dataset
sequence similarity-based data leakage between training, validation,
and test sets.

## Datasets

| Dataset | Task type | Labels | Metric | Provenance |
|---------|-----------|--------|--------|------------|
| `secstr` | per-residue classification | 3 | accuracy | DSSP assignments from [NetSurfP-2.0](https://doi.org/10.1002/prot.25674) (Klausen et al., 2019), test set from [ProtTrans](https://doi.org/10.1109/TPAMI.2021.3095381) NEW364 (Elnaggar et al., 2022) |
| `secstr8` | per-residue classification | 8 | accuracy | DSSP assignments from [NetSurfP-2.0](https://doi.org/10.1002/prot.25674) (Klausen et al., 2019), test set from [ProtTrans](https://doi.org/10.1109/TPAMI.2021.3095381) NEW364 (Elnaggar et al., 2022) |
| `disorder` | per-residue regression | continuous | spearman | [CheZOD database](https://doi.org/10.1038/s41598-020-71716-1) (Dass et al., 2020) -> [SETH](https://doi.org/10.3389/fbinf.2022.1019597) (Ilzhofer et al., 2022) -> [Schmirler et al., 2024](https://doi.org/10.1038/s41467-024-51844-2) |
| `rsa` | per-residue regression | continuous | spearman | Isolated-chain RSA from [NetSurfP-2.0](https://doi.org/10.1002/prot.25674) (Klausen et al., 2019), test set CB513 |
| `meltome` | per-protein regression | continuous | spearman | [Meltome Atlas](https://doi.org/10.1038/s41592-020-0801-4) (Jarzab et al., 2020) -> [FLIP](https://doi.org/10.1101/2021.11.09.467890) (Dallago et al., 2021) -> [Schmirler et al., 2024](https://doi.org/10.1038/s41467-024-51844-2) |
| `subloc` | per-protein classification | 10 | accuracy | [Light Attention](https://doi.org/10.1093/bioadv/vbab035) (Stärk et al., 2021) -> [Schmirler et al., 2024](https://doi.org/10.1038/s41467-024-51844-2) |
| `gpsite_dna` | per-residue classification | 2 | accuracy | [GPSite](https://doi.org/10.7554/eLife.93695) (Yuan et al., 2024) |
| `gpsite_rna` | per-residue classification | 2 | accuracy | [GPSite](https://doi.org/10.7554/eLife.93695) (Yuan et al., 2024) |
| `gpsite_pep` | per-residue classification | 2 | accuracy | [GPSite](https://doi.org/10.7554/eLife.93695) (Yuan et al., 2024) |
| `gpsite_pro` | per-residue classification | 2 | accuracy | [GPSite](https://doi.org/10.7554/eLife.93695) (Yuan et al., 2024) |
| `gpsite_atp` | per-residue classification | 2 | accuracy | [GPSite](https://doi.org/10.7554/eLife.93695) (Yuan et al., 2024) |
| `gpsite_hem` | per-residue classification | 2 | accuracy | [GPSite](https://doi.org/10.7554/eLife.93695) (Yuan et al., 2024) |
| `gpsite_zn` | per-residue classification | 2 | accuracy | [GPSite](https://doi.org/10.7554/eLife.93695) (Yuan et al., 2024) |
| `gpsite_ca` | per-residue classification | 2 | accuracy | [GPSite](https://doi.org/10.7554/eLife.93695) (Yuan et al., 2024) |
| `gpsite_mg` | per-residue classification | 2 | accuracy | [GPSite](https://doi.org/10.7554/eLife.93695) (Yuan et al., 2024) |
| `gpsite_mn` | per-residue classification | 2 | accuracy | [GPSite](https://doi.org/10.7554/eLife.93695) (Yuan et al., 2024) |

## Loading

```python
from datasets import load_dataset

# Load a single dataset (e.g. meltome)
ds = load_dataset("Moomboh/mutafitup-datasets", "meltome")

# Access splits
train = ds["train"]
valid = ds["validation"]
test = ds["test"]
```

## Resplit methodology

The original datasets were collected from three independent sources
(see provenance chain below). Because these sources share overlapping
protein sequences, naively combining their original train/test splits
introduces cross-dataset data leakage: a protein in one dataset's
training set may appear in another dataset's test set.

To eliminate this leakage, all sequences across all 16
datasets were pooled, deduplicated, and clustered using
**mmseqs** at **20.0% minimum sequence identity**.
New training, validation, and test splits were then assigned at the
cluster level so that no two splits share sequence-similar proteins:

1. **Cluster** all unique sequences across datasets using mmseqs
   at 20.0% minimum sequence identity.
2. **Merge** original train + valid pools per dataset.
3. **Remove** within-dataset test-similar sequences from the pool.
4. **Reassign** cross-dataset test-contaminated cluster members to
   validation.
5. **Top up** validation to at least 10.0% of the non-test data
   (random seed: 42).

**Shared protein groups:** (secstr, secstr8, rsa) --
datasets sharing the same underlying proteins are coordinated during
split assignment.

**Test set reconstruction:** `meltome` --
test sets are reconstructed from original test sequences with
cluster-aware decontamination.

**Duplicate aggregation:** `meltome` --
duplicate sequences within a split are aggregated (mean strategy for
regression scores).

The `metadata/` directory contains the full clustering and split
assignment artifacts for reproducibility.

## Parquet schemas

### Per-residue classification

Used by: `secstr`, `secstr8`, `gpsite_*`

| Column | Type | Description |
|--------|------|-------------|
| `sequence` | `str` | Amino acid sequence |
| `label` | `list[int]` | Per-residue integer class labels |
| `resolved` | `list[int]` | Per-residue binary mask (1 = ordered, 0 = disordered) |

### Per-residue regression

Used by: `disorder`, `rsa`

| Column | Type | Description |
|--------|------|-------------|
| `sequence` | `str` | Amino acid sequence |
| `score` | `list[float]` | Per-residue float scores (rsa uses 999.0 sentinel for unresolved residues) |

### Per-protein regression

Used by: `meltome`

| Column | Type | Description |
|--------|------|-------------|
| `sequence` | `str` | Amino acid sequence |
| `score` | `float` | Scalar score (melting temperature in degrees Celsius) |

### Per-protein classification

Used by: `subloc`

| Column | Type | Description |
|--------|------|-------------|
| `sequence` | `str` | Amino acid sequence |
| `label` | `str` | Class name (subcellular localization compartment) |
| `label_numeric` | `int` | Integer-encoded label |

## Data provenance

### Schmirler 2024 (meltome, subloc, disorder)

Training data from Schmirler, Heinzinger & Rost (2024), downloaded from
the companion GitHub repository
[RSchmirler/data-repo_plm-finetune-eval](https://github.com/RSchmirler/data-repo_plm-finetune-eval),
also archived at [Zenodo](https://doi.org/10.5281/zenodo.12770310).
Original train/valid/test splits were constructed by the authors using
MMseqs2 clustering at >20% pairwise sequence identity.

**Full provenance chains:**

- **meltome**: Experimental thermal proteome profiling from the
  [Meltome Atlas](https://doi.org/10.1038/s41592-020-0801-4)
  (Jarzab et al., 2020) -- standardized into benchmark splits by
  [FLIP](https://doi.org/10.1101/2021.11.09.467890)
  (Dallago et al., 2021) -- adopted with added validation set by
  [Schmirler et al., 2024](https://doi.org/10.1038/s41467-024-51844-2).
- **subloc**: Curated subcellular localization annotations from
  [Light Attention](https://doi.org/10.1093/bioadv/vbab035)
  (Starck et al., 2021) -- included directly by
  [Schmirler et al., 2024](https://doi.org/10.1038/s41467-024-51844-2).
- **disorder**: NMR-derived per-residue disorder Z-scores from the
  [CheZOD database / ODiNPred](https://doi.org/10.1038/s41598-020-71716-1)
  (Dass et al., 2020) -- used for pLM embedding prediction by
  [SETH](https://doi.org/10.3389/fbinf.2022.1019597)
  (Ilzhofer et al., 2022) -- re-split by
  [Schmirler et al., 2024](https://doi.org/10.1038/s41467-024-51844-2).

### NetSurfP-2.0 (secstr, secstr8, rsa)

Structure-derived annotations from
[NetSurfP-2.0](https://doi.org/10.1002/prot.25674)
(Klausen et al., 2019), hosted at DTU Health Tech. Secondary structure
labels are DSSP-based categorical assignments; RSA targets are
solvent-accessibility values for resolved residues.

The secondary-structure test set (NEW364) comes from the
[ProtTrans](https://doi.org/10.1109/TPAMI.2021.3095381)
project (Elnaggar et al., 2022). The RSA test set is CB513.

### GPSite (gpsite_*)

Residue-level binding-site annotations from
[GPSite](https://doi.org/10.7554/eLife.93695)
(Yuan et al., 2024), pinned to commit
[`58cfa4e`](https://github.com/biomed-AI/GPSite/tree/58cfa4e59f077e531fb38cf4a04bec6aea706454).
Labels correspond to experimentally resolved protein-ligand interactions
from the PDB across 10 ligand types (DNA, RNA, peptide, protein, ATP,
heme, Zn, Ca, Mg, Mn).

## References

- Dallago, C. et al. (2021). FLIP: Benchmark tasks in fitness landscape
  inference for proteins. *bioRxiv* 2021.11.09.467890.
  [doi:10.1101/2021.11.09.467890](https://doi.org/10.1101/2021.11.09.467890)
- Dass, R., Mulder, F.A.A. & Nielsen, J.T. (2020). ODiNPred:
  comprehensive prediction of protein order and disorder. *Sci. Rep.*
  10, 14780.
  [doi:10.1038/s41598-020-71716-1](https://doi.org/10.1038/s41598-020-71716-1)
- Elnaggar, A. et al. (2022). ProtTrans: Toward Understanding the
  Language of Life Through Self-Supervised Learning. *IEEE TPAMI*
  44(10), 7112--7127.
  [doi:10.1109/TPAMI.2021.3095381](https://doi.org/10.1109/TPAMI.2021.3095381)
- Ilzhofer, D., Heinzinger, M. & Rost, B. (2022). SETH predicts
  nuances of residue disorder from protein embeddings. *Front.
  Bioinform.* 2, 1019597.
  [doi:10.3389/fbinf.2022.1019597](https://doi.org/10.3389/fbinf.2022.1019597)
- Jarzab, A. et al. (2020). Meltome atlas -- thermal proteome stability
  across the tree of life. *Nat. Methods* 17, 495--503.
  [doi:10.1038/s41592-020-0801-4](https://doi.org/10.1038/s41592-020-0801-4)
- Klausen, M.S. et al. (2019). NetSurfP-2.0: Improved prediction of
  protein structural features by integrated deep learning. *Proteins*
  87(6), 520--527.
  [doi:10.1002/prot.25674](https://doi.org/10.1002/prot.25674)
- Schmirler, R., Heinzinger, M. & Rost, B. (2024). Fine-tuning protein
  language models boosts predictions across diverse tasks. *Nat.
  Commun.* 15, 7407.
  [doi:10.1038/s41467-024-51844-2](https://doi.org/10.1038/s41467-024-51844-2)
- Starck, H. et al. (2021). Light attention predicts protein location
  from the language of life. *Bioinformatics Advances* 1(1), vbab035.
  [doi:10.1093/bioadv/vbab035](https://doi.org/10.1093/bioadv/vbab035)
- Steinegger, M. & Soding, J. (2017). MMseqs2 enables sensitive protein
  sequence searching for the analysis of massive data sets. *Nat.
  Biotechnol.* 35, 1026--1028.
  [doi:10.1038/nbt.3988](https://doi.org/10.1038/nbt.3988)
- Yuan, Q., Tian, C. & Yang, Y. (2024). Genome-scale annotation of
  protein binding sites via language model and geometric deep learning.
  *eLife* 13, e93695.
  [doi:10.7554/eLife.93695](https://doi.org/10.7554/eLife.93695)