---
title: DeepSol troubleshooting
description: Why a DeepSol job rejects a sequence, times out or returns low confidence — ambiguous residues such as X or B, oversized batches, membrane proteins.
---

This page helps you diagnose the problems that most often surface with [DeepSol](/docs/deepsol) — invalid sequence characters, slow or timing-out batches, unexpectedly low confidence, and predictions that don't match experimental results.

## Common issues and fixes

## Invalid sequence errors

- Ensure sequences contain only standard amino acid characters (A, C, D, E, F, G, H, I, K, L, M, N, P, Q, R, S, T, V, W, Y).
- Remove any ambiguous residues (X, B, Z, J) or replace with the most likely amino acid.
- Check for numerical characters, spaces, or special symbols in your sequences.
- For FASTA uploads, verify header lines start with `>` and sequences are on subsequent lines.

## Timeouts or slow batches

- Reduce `batchSize` to 16 or smaller for very long protein sequences.
- Chunk large sequence libraries to under ~1,000 records per submission.
- Retry on `429` or `503` with exponential backoff and a small jitter.
- Consider splitting very long proteins (>2000 residues) into domains if biologically appropriate.

## Unexpectedly low confidence

- Check `warnings` for domain-fit notes; the CNN may be uncertain for unusual sequences.
- Verify sequence length is within the optimal range (50-2000 residues).
- Membrane proteins, intrinsically disordered regions, and unusual amino acid compositions may yield lower confidence.
- If the protein class is far from typical E. coli-expressed proteins, treat predictions as qualitative guidance only.

## Predictions don't match experimental results

- Remember DeepSol achieves 77% accuracy (MCC 0.55); some predictions will be incorrect.
- Expression conditions significantly affect actual solubility (temperature, IPTG concentration, media, strain).
- The model predicts intrinsic solubility tendency; fusion tags and chaperones can improve outcomes.
- Consider the experimental context: were inclusion bodies observed, or was protein simply not detected?

## Missing identifiers in outputs

- Include descriptive FASTA headers or an `id` column in CSV uploads; these are echoed back unchanged.
- When submitting multiple identical sequences, results will be deduplicated by default.

## Sequence formatting issues

- Remove line breaks within sequences (sequences should be continuous).
- Ensure consistent case (uppercase preferred, but lowercase is accepted).
- Strip any sequence numbering or annotation characters.
- Verify UTF-8 encoding for uploaded files.

## Still stuck?

- Capture the failing input sequences and job ID, then contact support with timestamps and workspace ID.
- Include whether you used FASTA, raw sequence, or CSV format and the chosen `batchSize`.
- Note any specific error messages or warnings returned by the system.

## Read next