Job Types Catalog
Searchable catalog of computational biology tools with categories and typical runtimes
The Job Types Catalog is the reference list of every computational tool you can run on the platform, grouped by category with its compute profile and typical runtime. Reach for it when you're deciding which tool fits your task or need a tool's exact name before submitting. Each entry can be invoked through the REST API, the SDK, or the Harness.
Difficulty
Time
Job Types Catalog
Complete catalog of computational tools available on the IvyAI platform. Each tool can be invoked via the REST API, SDK, or Harness.
Structure Prediction
| Tool | Description | Compute | Typical Runtime |
|---|---|---|---|
esmfold | ESMFold protein structure prediction | GPU | 30s–5min |
esm3_generate | ESM3 protein generation | GPU | 1–3min |
esm3_inverse_fold | ESM3 inverse folding (structure → sequence) | GPU | 30s–2min |
esm3_embeddings | ESM3 protein embeddings | GPU | 15–45s |
colabfold | ColabFold MSA-based structure prediction | GPU | 2–15min |
Molecular Docking & Design
| Tool | Description | Compute | Typical Runtime |
|---|---|---|---|
diffdock | DiffDock protein-ligand docking | GPU | 1–5min |
gnina | GNINA CNN-scored docking. Retired: use diffdock, then rescore with MM-GBSA | GPU | — |
colabdock | ColabDock flexible docking. Not currently callable: its AlphaFold2 model weights are missing | GPU | — |
smart_dock | SmartDock consensus docking | GPU | 3–8min |
reinvent4 | REINVENT4 de novo molecule generation | GPU | 5–30min |
genmol | GenMol molecular generation (SAFE native) | GPU | 2–10min |
denovo | De novo drug design | GPU | 5–20min |
Property Prediction
| Tool | Description | Compute | Typical Runtime |
|---|---|---|---|
admet_ai | ADMET property prediction | CPU | 10–30s |
admetlab3 | ADMETlab3 comprehensive ADMET profiling | CPU | 15–45s |
graphsol | GraphSol aqueous solubility prediction | CPU | 10–20s |
deepsol | DeepSol deep learning solubility. Not currently callable: disabled upstream (protobuf incompatibility); use graphsol | CPU | — |
protox3 | ProTox-3 toxicity prediction. Retired: use admet_ai for the overlapping endpoints | CPU | — |
deeppurpose | DeepPurpose drug-target interaction | GPU | 30s–2min |
Database Screening
| Tool | Description | Compute | Typical Runtime |
|---|---|---|---|
rdkit_full | RDKit SMILES validation and molecular properties | CPU | 5–15s |
zinc | ZINC database compound screening | CPU | 10–30s |
pubchem | PubChem compound queries | CPU | 5–15s |
pubchem_substructure | PubChem substructure search | CPU | 10–30s |
chembl | ChEMBL bioactivity database search | CPU | 10–30s |
Sequence Analysis
| Tool | Description | Compute | Typical Runtime |
|---|---|---|---|
blast | BLAST sequence homology search | CPU | 10s–2min |
biotite_fetch | UniProt/PDB structure fetching | CPU | 5–15s |
aizynthfinder | AiZynthFinder retrosynthetic route planning | CPU | 30s–5min |
Fragment & Scaffold Design
| Tool | Description | Compute | Typical Runtime |
|---|---|---|---|
fragment_growing | Fragment-based drug design. Retired: no replacement | GPU | — |
fragment_library | Fragment library management. Retired: no replacement | CPU | — |
fragment_search | Fragment discovery and search | CPU | 10–30s |
fragment_extract | Fragment extraction from structures | CPU | 10–30s |
scaffold_decoration | Scaffold decoration with R-groups | GPU | 2–10min |
linker_design | Linker design between fragments | GPU | 2–10min |
motif_extension | Motif-based molecule extension | GPU | 2–10min |
Literature & NLP
| Tool | Description | Compute | Typical Runtime |
|---|---|---|---|
biobert_pubmed | BioBERT scientific literature mining | GPU | 15–45s |
Protein Preparation
| Tool | Description | Compute | Typical Runtime |
|---|---|---|---|
pdb2pqr | Protonation state and charge assignment (PDB2PQR + PROPKA) | CPU | 15–30s |
p2rank | Binding pocket detection (P2Rank) | CPU | 30s–2min |
Protein preparation is these two job types, the same ones the chat interface runs (pdb2pqr_protonate, p2rank_predict). pdb2pqr takes inline structure text (pdb_content); p2rank takes pdb_content, a pdb_id or a uniprot_id. There are no proteinprep* job types, and no structure-quality job.
Regulatory & Safety
| Tool | Description | Compute | Typical Runtime |
|---|---|---|---|
drugcentral_regulatory | DrugCentral regulatory approvals lookup | CPU | 5–10s |
chembl_drug_warnings | ChEMBL drug warnings and withdrawals | CPU | 5–10s |
pubchem_regulatory | PubChem regulatory data (DEA, ATC) | CPU | 5–10s |
Genomics & Databases
| Tool | Description | Compute | Typical Runtime |
|---|---|---|---|
encode_search | ENCODE database functional genomics search | CPU | 10–30s |
string_network | STRING protein-protein interaction networks | CPU | 10–30s |
uniprot_search | UniProt protein database search | CPU | 5–15s |
pdb_search | RCSB PDB structure database search | CPU | 5–15s |
Analysis Levels
All tools support three analysis levels that control the depth of result processing:
| Level | Description |
|---|---|
basic | Fast results with standard output only |
standard | Balanced analysis with key quality metrics (default) |
comprehensive | Full analytical suite with extended result fields |
Example
Guest-Accessible Tools
Callers with no key reach an allowlisted guest catalog through the keyless orchestrate request, not by submitting a job type directly. See Guest curl for the request and its limit.
Programmatic Tool Discovery
Harness API
The web app serves the tool registry at /api/harness/tools:
SDK (planned)
@ivybiosciences/sdk is a planned specification, not a released package. Today, list tools with the REST calls above or ivybloom in the CLI.
Read next
- REST Jobs API — submit and manage jobs
- SDK Jobs Reference — TypeScript SDK job methods
- Harness Tools — filtered tool discovery for sessions
- Types & Constants — SDK type definitions
Was this page helpful?