Datasets:
metadata
license: cdla-permissive-2.0
task_categories:
- tabular-regression
pretty_name: true_betalactamase_complete
tags:
- tabular
- proteomics
- regression
- Biomedical
- Benchmark
- cds-seq
configs:
- config_name: true_betalactamase_complete
data_files:
- split: train
path: data/train.parquet
- split: validation
path: data/val.parquet
- split: test
path: data/test.parquet
dataset_info:
description: >-
Sequence-level regression task predicting the fitness of all single codon
substitutions in the TEM-1 Beta-lactamase gene. Labels measure the ability
to confer ampicillin resistance. The Complete dataset contains all coding
sequences (CDS) except those degenerate with any test-set CDS, using raw CDS
fitness values. Train and test sets are derived from Firnberg et al.
Features are mean-pooled ESM-2 (esm2_t12_35M_UR50D) sequence embeddings.