PyTorch-native graph neural networks for molecules and proteins. Use when building custom GNN architectures for drug discovery, protein modeling, or knowledge graph reasoning. Best for custom model development, protein property prediction, retrosynthesis. For pre-trained models and diverse featurizers use deepchem; for benchmark datasets use pytdc.
82
75%
Does it follow best practices?
Impact
98%
1.28xAverage score across 3 eval scenarios
Passed
No findings from the security scan
Fix and improve this skill with Tessl
tessl review fix ./skills/pharma/torchdrug/SKILL.mdRetrosynthesis two-stage pipeline
USPTO50k dataset
100%
100%
RGCN for center identification
100%
100%
GIN for synthon completion
0%
100%
CenterIdentification task
100%
100%
SynthonCompletion task
100%
100%
num_bond_type for RGCN
100%
100%
Two separate models
100%
100%
node_feature_dim from dataset
100%
100%
uv pip install
0%
0%
top_k for center identification
0%
100%
Knowledge graph completion for drug repurposing
RotatE or ComplEx model
100%
100%
Hetionet dataset
100%
100%
KnowledgeGraphCompletion task
100%
100%
adversarial_temperature param
100%
100%
num_negative param
100%
100%
num_entity from dataset
100%
100%
num_relation from dataset
100%
100%
BCE criterion for KG
100%
100%
uv pip install
0%
0%
Compound-disease query
100%
100%
Molecular GNN model selection and training
GIN model choice
100%
100%
Scaffold split
0%
100%
Node feature dim from dataset
0%
100%
Edge feature dim from dataset
0%
100%
Batch norm enabled
100%
100%
BCE criterion
100%
100%
AUROC metric
100%
100%
AUPRC metric
100%
100%
PropertyPrediction task
100%
100%
uv pip install
0%
100%
df37802
Table of Contents
If you maintain this skill, you can claim it as your own. Once claimed, you can manage eval scenarios, bundle related skills, attach documentation or rules, and ensure cross-agent compatibility.