eve-bio/drug-target-activity

Introduction

This dataset containing measurements of drug-target interactions is provided by EvE Bio. It is actively being generated with a quantitative screening process, and data for new targets is added every other month. For each target, one or more types of activity (agonism, antagonism, etc.) are measured for every drug in a 1,397 member compound library that primarily represents FDA approved small molecule drugs. Results are reported for every combination, whether they are active or inactive. Targets are members of three classes: nuclear receptors (NRs), 7-transmembrane receptors (7TMs, aka G-protein coupled receptors), and protein kinases. Data is also reported for cell viability, which is particularly important in conjunction with the 7TM results, which are generated with live cell assays (the other classes use biochemical assays). More information can be found on Eve's Data Page including methods detailing assays used and data processing. The data provided in this dataset, as well as the raw data, can be viewed interactively in the data explorer available on that site.

Since drugs are typically selective for a small number of targets, active results are sparse, on the order of 1%. The key response variables are compound activity and potency. Binary activity and maximum observed activity is captured for every compound-assay combination (outcome_is_active, outcome_max_activity). Activity is expressed as a % of maximum activity, in reference to known standard compounds for each assay. For active compounds that have sufficient potency to be measurable in the concentration range tested (is_quantified), four-parameter logistic curve fits result in quantified potency, measured as pXC50 (outcome_potency_pxc50). pXC50 is the negative log of the IC50/EC50 – the concentration at which half of the maximum activity is reached. Higher pXC50s are higher potency, and 5 is the lowest quantifiable pXC50 in the concentration range used.

To collect these measurements, EvE uses a two-phase quantitative screening process. All combinations of compounds and assays are included in the screening phase, which includes two replicates of three concentrations. A rules-based progression algorithm determines which compounds advance to the profiling phase, where the 11-point concentration range is 10 μM to 10 pM. The full concentration response is effectively censored by the concentration range tested. For low potency compounds, this leads to results that are reported as active, but not quantified.

In addition to cytotoxicity, compounds can interfere with assays in various ways, leading to potentially spurious results. Compounds that appear with suspicious frequency for any given target class and mode are flagged as “high frequency”. They could be removed from the data before model development, but in some cases true activity will be lost in the process. Alternatively, this frequency flag could be treated as a response in itself, in order to develop models that link compound and concentration response characteristics with particular forms of interference. Columns that flag combinations where either cell viability or hit frequency merit consideration are included in the dataset (viability_flag, frequency_flag).

The dataset contains one row per combination of target, compound, mode, and mechanism (currently there is only one mechanism per target class, but this will change when data for both signaling pathways is added for 7TMs in 2026). NRs and 7TMs have two modes each, while PKs and cell viability have one. Multiple identifiers are included for both compounds and targets. For compounds: SMILES (a text-based chemical representation), InChIkey, CAS #, UNII, and DrugBank ID. For targets: gene, Uniprot ID, and mutant/wildtype indicators.

Quickstart Guide

This guide demonstrates how to load and work with the drug-target activity dataset by target class, illustrating the elements of the dataset that are relevant by target class.

Setup: Load the Dataset

import polars as pl
from datasets import load_dataset

# Load dataset from Hugging Face
ds = load_dataset("eve-bio/drug-target-activity")
train_ds = ds['train']
df = pl.from_pandas(train_ds.to_pandas())

Nuclear Receptors (NR)

Nuclear receptors are addressed with a biochemical assay in two modes - Agonist and Antagonist. Since these are biochemical assays, cell viability is not relevant. High frequency compounds are flagged based on activity levels by mode (frequency_flag).

# Filter for nuclear receptor data
df_nr = df.filter(pl.col('target__class') == 'NR')

# Select relevant columns for NR analysis
df_nr_selected = df_nr.select([
    'target_id',
    'target__gene',
    'target__name',
    'target__uniprot_id',
    'compound_id',
    'compound__name',
    'compound__smiles',
    'mode',
    'mechanism',
    'outcome_is_active',
    'outcome_potency_pxc50',
    'outcome_max_activity',
    'is_quantified',
    'frequency_flag'
])

7TM Receptors (7TMs/GPCRs)

7TM receptors (G-protein coupled receptors) are addressed with cell-based assays in two modes - Agonist and Antagonist. The use of cell-based assays makes consideration of cell viability critical to data interpretation. Cell death can masquerade as antagonism. Compounds with potential viability concerns are flagged (viabiligy_flag) but direct comparison of potency and activity results for viability vs. 7TM results directly is necessary for proper interpretation of the data. Some compounds that are flagged for viability are still truly active. High frequency compounds are flagged based on activity levels by mode (frequency_flag). 7TMs can act through multiple mechanisms of action, and this is an important pharmacological consideration regarding biased agonism. These mechanisms are captured in mechanism (e.g., Barr2). Data for multiple mechanisms will be available starting in 2026.

# Filter for 7TM receptor data
df_7tm = df.filter(pl.col('target__class') == '7TM')

# Select relevant columns for 7TM analysis
df_7tm_selected = df_7tm.select([
    'target_id',
    'target__gene',
    'target__name',
    'target__uniprot_id',
    'compound_id',
    'compound__name',
    'compound__smiles',
    'mode',
    'mechanism',  # e.g., 'Barr2' for 7TM
    'outcome_is_active',
    'outcome_potency_pxc50',
    'outcome_max_activity',
    'viability_flag',  # Important: flags compounds that affect cell viability
    'is_quantified',
    'frequency_flag'
])

Protein Kinases

Protein Kinases are addressed with a biochemical assay in a single mode - Binding. Since these are biochemical assays, cell viability is not relevant. High frequency compounds are flagged based on activity levels by mode (frequency_flag). Protein Kinase targets include both wildtype and mutant variants. Use target__is_mutant and target__wildtype_id to analyze mutant selectivity.

# Filter for kinase data
df_kinase = df.filter(pl.col('target__class') == 'Kinase')

# Select relevant columns for kinase analysis
df_kinase_selected = df_kinase.select([
    'target_id',
    'target__gene',
    'target__name',
    'target__uniprot_id',
    'target__is_mutant',  # Boolean: is this a mutant kinase?
    'target__wildtype_id',  # Reference to wildtype target if mutant
    'compound_id',
    'compound__name',
    'compound__smiles',
    'mode',
    'outcome_is_active',
    'outcome_potency_pxc50',
    'outcome_max_activity',
    'is_quantified',
    'frequency_flag'
])

Cell Viability

For viability assays, outcome_is_active = True indicates the compound is potentially cytotoxic.

# Filter for viability assay data
df_viability = df.filter(pl.col('target__class') == 'Viability')

# Select relevant columns for viability analysis
df_viability_selected = df_viability.select([
    'target_id',
    'compound_id',
    'compound__name',
    'compound__smiles',
    'outcome_is_active',  # Active = potentially cytotoxic
    'outcome_potency_pxc50',
    'outcome_max_activity',
    'is_quantified'
])

Joining Viability Data to Other Target Classes

You can join viability data to 7TM data to identify compounds with potential cytotoxicity concerns. It is relevant to this target class due to the use of cell based assays, and is particularly critical for interpretation of antagonist data.

# Prepare viability outcomes for joining
viability_outcomes = df.filter(
    pl.col('target__class') == 'Viability'
).select([
    'compound_id',
    pl.col('outcome_is_active').alias('viability_is_active'),
    pl.col('is_quantified').alias('viability_is_quantified'),
    pl.col('outcome_potency_pxc50').alias('viability_pxc50'),
    pl.col('outcome_max_activity').alias('viability_max_activity')
])

# Join viability data to 7TM results
df_7tm_with_viability = df_7tm_selected.join(
    viability_outcomes,
    on='compound_id',
    how='left'  # Left join keeps all 7TM records, adds viability data where available
)

# Identify 7TM actives that may have viability concerns
potential_concerns = df_7tm_with_viability.filter(
    (pl.col('outcome_is_active') == True) &  # Active at 7TM target
    (pl.col('viability_flag') == True)  # Flagged for viability issues
)