> ## 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.
# Segmasker
> Segmasker measures the low-complexity content of protein sequences using the SEG algorithm. [Low-complexity regions](https://en.wikipedia.org/wiki/Low_complexity_regions_in_proteins) are stretches of biased amino acid composition, such as homopolymeric runs or short-period repeats, that can produce spurious matches during sequence comparison. For each input sequence, segmasker identifies the residues that fall within low-complexity regions and reports their count, their fraction of the sequence, and the sequence length, giving a quantitative measure of compositional bias.
**License:** Segmasker is licensed under Custom (NCBI BLAST+ public domain). Please refer to [the license](https://www.ncbi.nlm.nih.gov/IEB/ToolBox/CPP_DOC/lxr/source/scripts/projects/blast/LICENSE) for full terms.
Proto is not affiliated with NCBI. This toolkit is open source and builds on the implementation produced by this organization. Product names, logos, and trademarks are the property of their respective owners.
```bibtex theme={null}
@article{wootton1993seg,
title={Statistics of local complexity in amino acid sequences and sequence databases},
author={Wootton, John C and Federhen, Scott},
journal={Computers \& Chemistry},
volume={17},
number={2},
pages={149--163},
year={1993},
publisher={Elsevier},
doi={10.1016/0097-8485(93)85006-x}
}
@article{camacho2009blastplus,
title={BLAST+: architecture and applications},
author={Camacho, Christiam and Coulouris, George and Avagyan, Vahram and Ma, Ning and Papadopoulos, Jason and Bealer, Kevin and Madden, Thomas L},
journal={BMC Bioinformatics},
volume={10},
pages={421},
year={2009},
publisher={BioMed Central},
doi={10.1186/1471-2105-10-421}
}
```
| Function | Description | |
| ----------------- | ----------------------------------------------------------------------- | ---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
| `run_segmasker()` | Detect low-complexity regions in protein sequences using NCBI segmasker | Docs Source |
## Background
Most natural [proteins](https://en.wikipedia.org/wiki/Protein) contain regions whose [amino acid](https://en.wikipedia.org/wiki/Amino_acid) composition is strongly biased, including homopolymeric runs, short-period repeats, and segments dominated by a few residue types. These [low-complexity regions](https://en.wikipedia.org/wiki/Low_complexity_regions_in_proteins) are biologically real but cause difficulty in [sequence alignment](https://en.wikipedia.org/wiki/Sequence_alignment), because their similarity is driven by shared composition rather than by common ancestry, which inflates the apparent significance of matches between unrelated sequences.
The SEG algorithm ([Wootton and Federhen, 1993](https://doi.org/10.1016/0097-8485\(93\)85006-x)) quantifies local compositional complexity along a protein sequence using a sliding window and partitions the sequence into segments of low and high complexity. Masking or down-weighting the low-complexity segments before a similarity search improves the specificity of the results. Segmasker is the SEG implementation distributed as a command-line program within the NCBI [BLAST+](https://en.wikipedia.org/wiki/BLAST_\(biotechnology\)) suite ([Camacho et al., 2009](https://doi.org/10.1186/1471-2105-10-421)), which reorganized the original BLAST applications into modular command-line tools. Within that suite, segmasker applies the SEG procedure to protein sequences and reports the low-complexity regions it identifies, which can then be excluded from similarity searches or used to flag compositionally biased designs.
### Learning Resources
* [NCBI BLAST+ Command Line Applications User Manual](https://www.ncbi.nlm.nih.gov/books/NBK279690/) - the reference manual for the BLAST+ suite that segmasker ships with, including its masking applications.
* [BLAST Help (NCBI)](https://blast.ncbi.nlm.nih.gov/doc/blast-help/) - NCBI's documentation hub for BLAST concepts, including low-complexity filtering.
## Tools
### Segmasker Low-Complexity Detection (`segmasker-score`)
Applies the SEG algorithm to one or more protein sequences and returns, for each sequence, the number of residues classified as low-complexity, the fraction of the sequence those residues represent, and the sequence length. The low-complexity fraction is the primary metric for ranking sequences by compositional bias.
#### API Reference
Source
Protein sequence(s) to analyze for low-complexity regions. Can be provided as:
Source
Sliding-window size for SEG complexity analysis. Larger windows are less sensitive to short low-complexity stretches.
Lower complexity cutoff. Regions scoring below this are classified as low-complexity.
Upper complexity cutoff. Defines the transition between masked and unmasked regions. Must be >= `locut`.
Verbosity level (0=quiet, 1=info, 2=debug, 3=raw subprocess stderr). `True` is coerced to `1` and `False` to `0`.
Device to run the tool on.
Maximum execution time in seconds. `None` waits indefinitely.
Random seed. When set, tools run reproducibly up to small GPU float noise (see `BaseToolOutput.approx_equal`), and the seed participates in cache keys. When None, cacheable seed-sensitive tools skip cache until seeded.
Source
Per-sequence low-complexity metrics, index-aligned with `inputs.sequences`.
Name of the metric that best summarizes the result overall (e.g. `"avg_plddt"` for AlphaFold2). Used by downstream UI and reporting to pick a headline value.
Concrete Metrics subclass tag; enables typed reconstruction after a serialization round-trip.
**Metrics**
| Metric | Type | Range | Availability |
| ------------------------- | ----- | ---------- | ------------ |
| `low_complexity_fraction` | float | 0.0 to 1.0 | always |
| `low_complexity_count` | int | ≥ 0.0 | always |
| `sequence_length` | int | ≥ 1.0 | always |
#### Applications
* Screening designed protein sequences for compositional bias before further analysis.
* Quantifying low-complexity content to flag homopolymeric runs or short-period repeats.
* Prioritizing sequences for masking ahead of a protein similarity search to reduce spurious matches.
#### Usage Tips
* **`window` sets the scale of the regions detected.** A larger window targets broader low-complexity stretches, while a smaller window resolves shorter runs.
* **`locut` and `hicut` set how aggressively regions are flagged.** Raising the cutoffs classifies more of the sequence as low-complexity, while lowering them applies a stricter criterion that flags only the most biased regions. `hicut` must be greater than or equal to `locut`.
* **Very short and empty sequences are limited.** A sequence shorter than the window cannot be assessed reliably, and an empty sequence reports a low-complexity fraction of zero.
## Toolkit Notes
* **Detection runs on CPU and is deterministic.** Segmasker takes only protein sequences, runs without a GPU, and returns the same values for identical inputs on repeated calls.
* **Results are index-aligned with the input.** Each result corresponds to the input sequence at the same position, so a batch of sequences returns metrics in the order they were supplied.
**Example notebook:** See the [full working example](https://github.com/evo-design/proto-tools/blob/47e34afa5ea240a3b406e323dc38aa5dc85f223e/proto_tools/tools/sequence_scoring/segmasker/examples/example.ipynb) for a copy-paste-ready walkthrough.
## Infrastructure Guides
The following guides cover how to run tools efficiently and at scale.
Keep a tool's model warm across calls instead of reloading it every invocation.How GPUs are allocated to tools and how to target specific devices.Fan a batch of inputs out across multiple GPUs.
evo-design/proto-tools/proto\_tools/tools/sequence\_scoring/segmasker
NCBI
