Apheris Hub - Known Issues

Below, we list known issues with Apheris Hub. If available, we list a workaround to help troubleshoot the issue.

This list will be constantly evolving with input from our different users and with the evolution of our product in future releases.

Known Issues with Apheris Hub 1.3

OpenFold3

Issue	Description	Additional information
Unexpected Behavior on some Batch Requests	Due to an unresolved MSA cache bug in the OpenFold3 code base, batch requests (i.e. requests with more than one query) with more than a single unique protein sequence and counterpart MSA may cause unexpected behavior.	Please submit individual requests with a single query for these complexes.
MSA Files Cannot Be Used with OpenFold3	An issue prevents the reliable usage of MSA files with the current OpenFold3 model.	Do not use MSA files with the OpenFold3 model version. The public MSA server version is available, but you should not use MSA files with it.

Prediction and Processing

Issue	Description	Additional information
ColabFold Server Rate Limiting	The public ColabFold server applies rate limiting, which can significantly slow down processing of even simple queries.	If you encounter slower than anticipated query times it is likely due to using the public ColabFold server. Please wait a few minutes and try again or use MSAs from file if the model supports it.
Delayed Execution on Shared GPU	When we have a single GPU being shared between two running model versions, parallel jobs can be started on each of those models. However, because they share the same GPU, the execution can be blocked and delayed.	Additionally, the UI shows the delayed requests as "Running" instead of "Pending", as they are waiting for the GPU to become available.
Error Log Display	Error logs are shown even for successful prediction runs. However, they don't always contain errors per se but just the stderr output.
Job Duration Metrics	Job duration doesn't distinguish between waiting time and actual running time, which can be misleading when multiple jobs are queued.
No validation that MSA file sequence and sequence provided in queries match	Queries containing unknown residues (X character) at the beginning of sequences, particularly when combined with ligands like HEM, may cause MSA generation errors.	Future releases will improve the validations being performed before a user submitting a prediction run known to fail.
Nonstandard residues between parentheses are not supported	Including nonstandard residues in parentheses causes the MSA to crash. Generally, these should be converted to X before sending to an MSA server.	Future releases will add support for Nonstandard residues.
MSA Server Timeout Ignored When Not Less Than Polling Interval	When setting an MSA server timeout that is not less than the MSA polling interval, jobs may fail with an "invalid response from provider" error instead of a proper timeout error.	Ensure the MSA server request timeout is set to a value less than the MSA polling interval.

Benchmarking

Issue	Description	Additional information
Boltz-2 and Protenix Benchmarks May Omit Ligand Metrics	Benchmark runs using Boltz-2 or Protenix may sometimes omit ligand metrics in the results, particularly for predictions with less accurate atom distances.
Benchmark Job Metrics May Not Load Automatically	When viewing a benchmark's results page while the benchmark is still running, the metrics for at least one job may not appear automatically once the benchmark completes.	Refresh the page to load the missing metrics.

Fine-Tuning

Issue	Description	Additional information
Dataset Preparation Fails with Non-Numeric Chain IDs When Using MSA Service	When relying on the MSA service for automatic MSA generation (not providing custom MSA files), structures with non-numeric chain IDs (e.g., "1.A") cause dataset preparation to fail. The fine-tuning status moves from "validating" back to "created" with the error "Dataset generation failed while waiting for the job to start." Logs show an error about missing MSA paths for protein chains with the unsupported chain IDs.	Either upload MSA files manually for affected structures, or modify the CIF structure to use numeric-only chain IDs (e.g., "1", "2" instead of "1.A"). Numeric chain IDs work correctly with the MSA service.
Download Checkpoint Button Disabled for Last Checkpoint	After completing an OpenFold3 fine-tuning run, the download button for the last checkpoint is always disabled, even though the checkpoint file exists in the output folder.	This affects only the last checkpoint; earlier checkpoints can be downloaded normally.
OpenFold3 model running both inference and fine-tuning capabilities required to run fine-tuning	In order to run fine-tuning, the deployed Hub should have access to an OpenFold3 model running both inference and fine-tuning capabilities.
Dataset Generation May Fail Intermittently for Jobs with SDF Files	Fine-tuning jobs that include SDF files may sometimes fail during the dataset generation (validation) phase. This is an intermittent issue that can occur with the currently supported SDF structure format.	Retry the fine-tuning job if dataset generation fails.
Dataset Generation Fails for SDF Files with Explicit Hydrogens	Fine-tuning dataset generation fails when SDF files contain explicit hydrogen atoms. SDF files with explicit hydrogens are not currently supported and will cause the dataset generation phase to fail.	Remove explicit hydrogens from SDF files before uploading them for fine-tuning. Tools such as RDKit or Open Babel can be used to strip hydrogens from SDF files (e.g., using RDKit's RemoveHs() or Open Babel's -d option).

Security

Issue	Description	Additional information
Container User Privileges	Docker containers are currently running with root user privileges, which is not a security best practice.

User Interface

Issue	Description	Additional information
PAE Filter Issues on Window Resize	When resizing the browser window or zooming out while viewing results, the PAE filter selection may appear and disappear unexpectedly. Clicking deselect may not work consistently.	If you experience this issue, try avoiding window resize operations while working with PAE plots.
Model and Weights Search Filter Not Working Correctly	The search filter for models and weights on the Predict and Benchmark pages does not filter results consistently. Typing more characters may unexpectedly show more results instead of narrowing them down.
Model Queue Overlaps Fullscreen Viewer Close Button	When jobs are queued for one or more models, the model queue notification overlaps the close button on the fullscreen molecule/matrix viewer, preventing it from being clicked.	Use the "disable fullscreen" button as an alternative to close the fullscreen viewer.
Unclear Error Feedback for Invalid Structure Files	When using an invalid or broken CIF file to start a prediction via "Start from CIF", the error message shown is not descriptive enough to inform the user that the structure file is invalid.
MSA Progress Bar Not Shown in Additional Browser Tabs	When MSA preparation for a prediction or benchmark is running, opening the results list in another browser tab shows "pending" or "preparing" status instead of the MSA progress bar.	The MSA progress bar is displayed correctly in the original tab. Use the original tab to monitor MSA preparation progress.
Cannot Switch Users After Logout	After logging out, logging back in with the same organization automatically signs in as the previous user. It is not possible to select a different user without clearing the browser cache or using a new window.	Clear browser cache or use an incognito/private window to log in as a different user within the same organization.
API Documentation Error for Support Endpoint	The API documentation shows an error when viewing the GET /api/v1/support/zip endpoint.
API Documentation Formatting Issues on Fine-Tune Endpoints	The API documentation for fine-tune endpoints shows misaligned path parameter columns and a schema error on the GET /api/v1/fine-tune/{id}/output/{key} endpoint.

Form and Input Handling

Issue	Description	Additional information
Duplicate Structure Upload Not Indicated in Benchmarks	When setting up a benchmark, uploading a structure file that has already been added does not show any warning or indication that the file is a duplicate.
Invalid CIF Upload Accepted Without Error Status	When uploading an invalid CIF file to a fine-tune dataset, the API returns a success status instead of an error. The frontend works around this by checking the response body, but the API response code is misleading.
Fine-Tune Job Form Retains Previous Job's Title and Description	When clicking on a new fine-tuning job without refreshing the browser, the title and description fields retain the values from the previously created job instead of being cleared.	Refresh the browser before creating a new fine-tuning job.
API Error When Removing Structure from Fine-Tuning Job	Navigating back to a previously created fine-tuning job and attempting to remove a structure file results in an API error.
Re-uploading Deleted Structure File Does Nothing	When removing an uploaded structure file on the fine-tuning form and attempting to re-upload the same file, nothing happens. The file is not added back.	Refresh the browser or click cancel before re-uploading the same structure file.

Installation and Configuration

Issue	Description	Additional information
Port Allocation	The installation script uses a fixed port (8080) which may already be in use on some systems.	As a workaround, ensure port 8080 is available before installation or modify the port in the installation script.
Single GPU Limitation with Multiple Models	When deploying the Hub using Kubernetes and Helm on a machine with only one GPU, only one model (OpenFold3 or Boltz-2) can be started at a time.	Use a machine with multiple GPUs, enable fractional GPU usage, or run one model at a time. This limitation should be considered when planning your deployment infrastructure.
Bash Version Check Missing in Deploy Script	The deploy_apherisfold script requires GNU Bash version 4.0 or higher but does not check this requirement before execution. Running the script with an older version or a different shell results in cryptic error messages.	Ensure you are using GNU Bash version 4.0 or higher before running the deployment script. Verify your bash version with bash --version.