> ## Documentation Index
> Fetch the complete documentation index at: https://proto.evodesign.org/docs/llms.txt
> Use this file to discover all available pages before exploring further.
# Overview
> How proto-tools works: the universal tool pattern, categories, and execution model
# Overview
proto-tools provides standardized Python implementations of 60+ bioinformatics tools: structure predictors, sequence scorers, gene annotators, alignment engines, and more. Every tool follows the same Input / Config / Output pattern, so learning one tool transfers to the rest.
## The Input / Config / Output Pattern
Every tool follows a three-part pattern using Pydantic models:
```mermaid theme={null}
flowchart LR
I["ToolInput
(primary data)"] --> T["Tool Function
run_tool()"]
C["ToolConfig
(parameters)"] --> T
T --> O["ToolOutput
(results + metadata)"]
classDef default fill:#5D6F75,stroke:none,color:#FFFFFF,rx:6,ry:6
classDef primary fill:#046E7A,stroke:none,color:#FFFFFF,rx:6,ry:6
class T primary
```
```mermaid theme={null}
flowchart LR
I["ToolInput
(primary data)"] --> T["Tool Function
run_tool()"]
C["ToolConfig
(parameters)"] --> T
T --> O["ToolOutput
(results + metadata)"]
classDef default fill:#344649,stroke:none,color:#FFFFFF,rx:6,ry:6
classDef primary fill:#0A7E8C,stroke:none,color:#FFFFFF,rx:6,ry:6
class T primary
```
**Input**: *What* to analyze. The primary data: sequences, structures, files.
```python python icon="python" theme={null}
ESMFoldInput(
complexes=[
Complex(
chains=[
Chain(sequence="MKTAYLLIGL...")
]
)
]
)
```
**Config**: *How* to analyze it. Parameters and settings. Always optional; sensible defaults are built in.
```python python icon="python" theme={null}
ESMFoldConfig(
num_recycles=4,
)
```
**Output**: Results plus standardized metadata (execution time, success status, tool ID, warnings):
```python python icon="python" theme={null}
output = run_esmfold(input, config)
output.success # True
output.execution_time # 12.3 (seconds)
output.tool_id # "esmfold-prediction"
output.structures[0] # Structure object with coordinates, pLDDT, etc.
```
## Tool Categories
Predict 3D structures from sequences.
**AlphaFold2**, **AlphaFold3**, **Boltz2**, **Chai1**, **ESMFold**, **ESMFold2**, **Protenix**, **ViennaRNA**
Generate novel protein backbone structures.
**RFDiffusion3**
Sample conformational ensembles.
**BioEmu**
Design sequences for target structures.
**ProteinMPNN**, **LigandMPNN**, **FAMPNN**, **ESM-IF1**
Protein language models for scoring and sampling.
**ESM2**, **ESM3**, **ESMC**, **AbLang**
Autoregressive models for generation and scoring.
**Evo1**, **Evo2**, **ProGen2**, **ProGen3**
Predict functional effects from genomic sequences.
**Enformer**, **Borzoi**, **AlphaGenome**, **Malinois**, **Puffin**, **Segmasker**
Annotate sequences and find functional elements.
**PyHMMER**, **CRISPRtracrRNA**, **MinCED**, **Promoter Calculator**
Align sequences and search databases for homologs.
**BLAST**, **MMseqs2**, **MAFFT**, **ColabFold Search**
Find open reading frames in DNA.
**Orfipy**, **Prodigal**
Predict splice sites and specificity.
**SpliceTransformer**, **Pangolin**, **SpliceAI**
Align and compare 3D structures.
**TMAlign**, **USAlign**, **Foldseek**, **FoldMason**, **PyMOL RMSD**
Fetch sequences and structures from public databases.
**UniProt**, **PDB**, **NCBI**, **SequenceFetch**
Score and analyze 3D structure quality.
**DSSP**, **IPSAE**, **pDockQ2**, **PyRosetta**, **Structure Metrics**
De novo antibody and binder design pipelines.
**BindCraft**, **Germinal**
Random sequence mutagenesis.
**Random Protein**, **Random Nucleotide**
## GPU vs CPU Tools
### GPU Tools
Deep learning models that require NVIDIA GPUs. Faster but require specific hardware.
* **Structure Prediction:** AlphaFold3, Boltz2, Chai1, ESMFold, Protenix
* **Inverse Folding:** ProteinMPNN, LigandMPNN, FAMPNN
* **Language Models:** ESM2, ESM3, ESMC, AbLang, Evo1, Evo2, ProGen2, ProGen3
* **Sequence Scoring:** Enformer, Borzoi, AlphaGenome
* **Structure Design:** RFDiffusion3
* **Structure Dynamics:** BioEmu
* **RNA Splicing:** SpliceTransformer
### CPU Tools
Classical bioinformatics algorithms and binary tools. Run anywhere.
* **Gene Annotation:** PyHMMER, CRISPRtracrRNA, MinCED, Promoter Calculator
* **Sequence Alignment:** BLAST, MMseqs2, MAFFT, ColabFold Search
* **ORF Prediction:** Orfipy, Prodigal
* **Structure Prediction:** ViennaRNA (RNA only)
* **Structure Alignment:** TMAlign, USAlign
* **Database Retrieval:** UniProt, PDB, NCBI
## Environment Isolation
Some tools have complex or conflicting dependencies. These tools use isolated virtual environments managed by `ToolInstance`:
* Each tool with isolated deps has a `standalone/` directory with `setup.sh` and `run.py`
* Virtual environments are created automatically on first use
* Execution is handled transparently; you call the same `run_tool()` API
See the [Tool Environments](/docs/tools/guides/tool-environments) guide for details on how this works.
## Tool Registry
All tools are registered via the `@tool()` decorator, enabling automatic discovery and schema generation:
```python python icon="python" theme={null}
from proto_tools.tools.tool_registry import ToolRegistry
# List all available tools
all_tools = ToolRegistry.list_all()
for tool_spec in all_tools:
print(f"{tool_spec.key}: {tool_spec.label}")
# Get a specific tool's schema
schema = ToolRegistry.get_schemas("esmfold-prediction")
# Get a minimal example input
example = ToolRegistry.get_example_input("esmfold-prediction")
# Get citation
citation = ToolRegistry.get_citation("esmfold-prediction")
```
## Next Steps
Structure and Ligand data objects used by tools
Run your first tool in 5 minutes
Batch workloads with persistent tool instances
GPU allocation and multi-device execution