metadata

license: mit
configs:
  - config_name: dataset
    data_files: dataset.csv

Description

Fluorescence prediction is a regression task where each input protein x is mapped to a label y ∈ R, corresponding to the log-fluorescence intensity of x.

Splits

Structure type: None

The dataset is from Evaluating Protein Transfer Learning with TAPE. We follow the original data splits, with the number of training, validation and test set shown below:

Train: 20963
Valid: 5235
Test: 25517

Label

The label means the log-fluorescence intensity of the sequence.