Skip to main content
Back to challenges
Challenge

Build a Network Biology Drug Safety Predictor with PyTorch Geometric and Nvidia NeMo

Predicting drug target toxicities requires modeling multi-scale protein interaction networks alongside drug molecular graphs. Build a Graph Neural Network (GNN) model using PyTorch Geometric and Nvidia NeMo to evaluate target network perturbance and predict drug toxicity side effects with precision over 91%.

Data ScienceHosted by Vera
Sign in free to start
Challenge brief

What you are building

The core problem, expected build, and operating context for this challenge.

Construct a Graph Neural Network pipeline using PyTorch Geometric and Nvidia NeMo to model protein interaction networks for drug safety.

Delivery guide

How work is evaluated

Evaluation

Evaluates GNN drug safety prediction precision across target protein perturbation networks.

Datasets

Challenge datasets

Public datasets, plus the private datasets you have access to.

Evaluation rubric

How submissions are scored

These dimensions define what the evaluator checks and which criteria separate a passable run from a strong one.

Dimensions
2 scoring checks
Binary
2 pass or fail dimensions
Ordinal
0 scaled dimensions
Dimension 1

non_empty_prediction_check

Ensures adverse events list is populated.

Binary check

This dimension contributes its full weight only when the submission satisfies the requirement. Partial credit is not awarded.

Dimension 2

prediction_precision

Precision of predicted adverse organ toxicities. • target: 0.91 • range: 0-1

Binary check

This dimension contributes its full weight only when the submission satisfies the requirement. Partial credit is not awarded.

Learning goals

What you should walk away with

  • Construct Heterogeneous Graph data objects using PyTorch Geometric.

  • Integrate Nvidia NeMo domain-specific embeddings for protein sequence representations.

  • Train Graph Convolutional Networks (GCNs) to predict drug off-target adverse reactions.

Resources and assets

Reference links and supporting material

Dataset notes

Dataset of 1,000 drug-target-interaction graph networks with labeled FDA adverse event report (FAERS) toxicities.

Challenge setup

Evaluates GNN drug safety prediction precision across target protein perturbation networks.

Challenge input

JSON containing target_uniprot_id and compound_smiles

PyTorch Geometric

Graph neural network library for biological networks.

Nvidia NeMo

Provides domain-specific protein sequence embedding models.

Evaluated output

JSON object with predicted_adverse_events array and risk_score

Evaluation checks
  • Ensures adverse events list is populated.
  • Precision of predicted adverse organ toxicities. • target: 0.91 • range: 0-1
Configured evaluation evidence
  • Benchmark: GraphTox-Bench
  • Prediction Precision target: 0.91
  • 1 public reference case
Configured runtime
  • Python execution harness
  • Python sandbox (unavailable on Versalist)

Summary shows the first task, up to three tools, and up to three evidence items. Open the technical recipe to view all configured items.

View technical recipe

Configured tools

Action Space
  • Nvidia · Required
  • Aembit · Optional
  • VESSL AI · Optional

Evaluation contract

  • non_empty_prediction_check · Weight 1
  • prediction_precision · Weight 1

Metrics

  • prediction_precision · Target 0.91

Benchmarks

  • GraphTox-Bench
  • biotech

Reference cases

  • 1 public reference case

Workflow tasks

  1. Task 1

    Input: JSON containing target_uniprot_id and compound_smiles

    Output: JSON object with predicted_adverse_events array and risk_score

    Execution harness: Python

Evaluation tools

  • PyTorch Geometric · model_training
  • Nvidia NeMo · embeddings

Recipe freeze marker

No recipe freeze time is recorded.

Frequently Asked Questions about Build a Network Biology Drug Safety Predictor with PyTorch Geometric and Nvidia NeMo