> ## Documentation Index
> Fetch the complete documentation index at: https://docs.openbio.tech/llms.txt
> Use this file to discover all available pages before exploring further.

# Predictions

> Submit and monitor AI-powered prediction jobs

Submit AI-powered prediction jobs through OpenBio to analyze protein structures, design new molecules, and run complex computational biology analyses. Monitor your jobs in real-time and view interactive results directly in your workspace.

<Info>
  **What are Predictions?**

  Predictions are computational jobs that use state-of-the-art AI or traditional computational models and tools to analyze biological data. OpenBio handles all the technical infrastructure—you just describe what you want to analyze.
</Info>

## Available Prediction Tools

### Structure Prediction

<CardGroup cols={2}>
  <Card title="BoltzGen">
    **Purpose**: De novo protein backbone and sequence generation

    **Input**: Target specifications or structural constraints

    **Output**: Generated protein backbones and sequences
  </Card>

  <Card title="Boltz-2">
    **Purpose**: Protein-ligand co-folding

    **Input**: Protein sequence + optional ligand structure

    **Output**: Protein-ligand complex structure
  </Card>

  <Card title="Chai-1">
    **Purpose**: Multi-modal foundation model for molecular structure prediction

    **Input**: Proteins, DNA, RNA, ligands, glycans

    **Output**: Molecular structure
  </Card>

  <Card title="SimpleFold">
    **Purpose**: Fast structure prediction

    **Input**: Protein sequence only

    **Output**: Protein structure
  </Card>
</CardGroup>

### Protein Design

<CardGroup cols={2}>
  <Card title="ProteinMPNN">
    **Purpose**: Design sequences for structures

    **Input**: Protein backbone structure (PDB)

    **Output**: Optimized amino acid sequences
  </Card>

  <Card title="LigandMPNN">
    **Purpose**: Design protein binders

    **Input**: Target protein + binding site

    **Output**: Potential binder sequences
  </Card>

  <Card title="ThermoMPNN">
    **Purpose**: Improve protein stability

    **Input**: Protein structure

    **Output**: Thermostable variants
  </Card>

  <Card title="Pinal">
    **Purpose**: Design from text description

    **Input**: Natural language description

    **Output**: Designed protein structure
  </Card>
</CardGroup>

### Docking & Analysis

<CardGroup cols={1}>
  <Card title="GeoDock">
    **Purpose**: Protein-protein docking

    **Input**: Proteins (PDB)

    **Output**: Protein-protein complex structure
  </Card>
</CardGroup>

<Tip>You can always ask OpenBio to tell you more about a prediction tool.</Tip>

## Submitting Prediction Jobs

### Through Chat (Recommended)

The most natural way to submit predictions is through conversation:

<Tabs>
  <Tab title="Structure Prediction">
    ```
    Predict the structure of this protein using Boltz-2:
    MKTAYIAKQRQISFVKSHFSRQLEERLGLIEVQAPILSRVGDGTQDNLSGAEKAVQVKVKALPDAQFEVVHSLAKWKRQTLGQHDFSAGEGLYTHMKALRPDEDRLSPLHSVYVDQWDWERVMGDGERQFSTLKSTVEAIWAGIKATEAAVSEEFGLAPFLPDQIHFVHSQELLSRYPDLDAKGRERAIAKDLGAVFLVGIGGKLSDGHRHDVRAPDYDDWSTPSELGHAGLNGDILVWNPVLEDAFELSSMGIRVDADTLKHQLALTGDEDRLELEWHQALLRGEMPQTIGGGIGQSRLTMLLLQLPHIGQVQAGVWPAAVRESVPSLL
    ```
  </Tab>

  <Tab title="Using Files">
    ```
    Predict the structure of the protein in @my_sequence.fasta using Chai-1
    ```
  </Tab>

  <Tab title="Protein Design">
    ```
    Design a thermostable variant of this protein structure using ThermoMPNN: @enzyme.pdb
    ```
  </Tab>

  <Tab title="With workflow">
    ```
    Run /structure-prediction using @my_protein.fasta
    ```
  </Tab>
</Tabs>

### What Happens Next

<Steps>
  <Step title="AI Understands Your Request">
    OpenBio analyzes your message and selects the appropriate prediction tool
  </Step>

  <Step title="Parameter Confirmation">
    The AI may ask for clarification or confirm parameters before submitting
  </Step>

  <Step title="Job Submission">
    Your prediction is queued and assigned a unique job ID
  </Step>

  <Step title="Real-time Updates">
    Monitor progress through status updates in your chat
  </Step>

  <Step title="Results Delivery">
    When complete, results automatically appear in your research panel
  </Step>
</Steps>

### Through the Predictions Page

The support for submitting predictions through the predictions page is coming soon.

## Monitoring Your Predictions

### Job Status Tracking

<Info>
  **Location**: Research panel → Predictions tab, or click "Predictions" in top navigation
</Info>

The jobs will show up in the predictions page. They can have the following statuses:
**1. Queued**: Waiting in line to start
**2. Running**: Actively processing on GPU servers
**3. Completed**: Successfully finished
**4. Failed**: Error occurred during processing

### Viewing Job Details

Click on any prediction job to see comprehensive information:

* **Job ID**: Unique identifier for reference
* **Tool and Version**: Which AI model was used
* **Submission Time**: When job was submitted
* **Runtime**: How long the job has been running
* **Input Parameters**: What data was provided
* **Error Logs**: For failed jobs (if applicable)

<Check>
  You can refresh the page to see the latest status of your jobs.
</Check>

## Working with Results

The results are usually files that are automatically saved to your project's filesystem. You can see them from the files tab or ask OpenBio to tell you more about them.

<Tip>You can download the results from the files tab.</Tip>

## Retrying Failed Jobs

If a prediction fails:

1. **Review Error Message**: Check what went wrong
2. **Fix Input**: Address the issue (format, missing data, etc.)
3. **Resubmit**: Click "Retry" or create a new job
4. **Monitor**: Watch the new job progress

<Tip>OpenBio can help you troubleshoot the issue and retry the job for you.</Tip>

## Best Practices

<AccordionGroup>
  <Accordion title="Validate Input First">
    Before submitting:

    * Check sequence format (no invalid characters)
    * Verify file types are supported
    * Confirm required files are uploaded
  </Accordion>

  <Check>
    OpenBio automatically validates the input and always asks you to confirm the parameters before submitting the job.
  </Check>

  <Accordion title="Choose the Right Tool">
    * **Fast prediction needed** → SimpleFold
    * **Highest accuracy** → Boltz-2 or Chai-1
    * **Multiple molecule types** → Chai-1
    * **Protein design** → BoltzGen or ProteinMPNN family
    * **Protein-protein docking** → GeoDock
  </Accordion>

  <Accordion title="Be Patient">
    Structure predictions take time. Don't submit duplicates.
  </Accordion>

  <Accordion title="Monitor Progress">
    Keep an eye on running jobs. You can see the progress in the predictions page.
  </Accordion>

  <Accordion title="Download Important Results">
    Always download critical results:

    * Local backup of important structures
    * Save analysis reports for documentation
    * Keep sequences and scores for reference
  </Accordion>
</AccordionGroup>

## Troubleshooting

<AccordionGroup>
  <Accordion title="Immediate Failure">
    **Common causes**:

    * Invalid sequence format
    * File size too large
    * Missing required parameters
    * Unsupported file type

    **Solution**: Check error message, fix input, retry
  </Accordion>

  <Accordion title="Results Look Wrong">
    **Check**:

    * Confidence scores (low = unreliable) where applicable
    * Input sequence was correct
    * Try different prediction tool
    * Compare with known structures if available
  </Accordion>

  <Accordion title="Can't Find Results">
    **Solutions**:

    * Look on Predictions page for completed jobs
    * Refresh the page if needed
    * Check the files tab for the results
    * Ask OpenBio to tell you more about the results
  </Accordion>
</AccordionGroup>

## Next Steps

<CardGroup cols={2}>
  <Card title="Visualizations" href="/user-guide/features/visualizations">
    Analyze and interact with prediction results
  </Card>

  <Card title="Workflows" href="/user-guide/features/workflows">
    Create automated prediction workflows
  </Card>

  <Card title="File Management" href="/user-guide/features/file-management">
    Organize prediction inputs and outputs
  </Card>

  <Card title="Tutorials" href="/user-guide/tutorials/protein-structure-prediction">
    Step-by-step prediction tutorials
  </Card>
</CardGroup>