🧬 FastSTR

FastSTR — Ultra-fast and accurate identification of Short Tandem Repeats (STRs) from long-read DNA sequences. Developed for genome-wide STR detection, consensus construction, and comparative STR analysis.

---

📘 Table of Contents

1. Overview
2. Installation
3. Quick Start
4. Command Line Options
5. Input & Output
6. Usage
7. Performance
8. Citation
9. License
10. Changelog

---

🌍 Overview

FastSTR is a novel and efficient tool for de novo detection of short tandem repeats (STRs) in genomic sequences. It combines fast motif recognition with accurate sequence alignment to achieve both high precision and completeness in STR identification. FastSTR is optimized for large-scale genomic datasets and enables rapid detection of repetitive elements without relying on predefined motif libraries or fixed repeat-length thresholds.

Compared to classical tools like TRF, T-reks, and TRASH, FastSTR achieves:

• ⚡ High-speed parallel processing — Processes genomic fragments in parallel, achieving up to 10× faster runtime.
• 🧠 Context-aware motif recognition — Uses an N-gram + Markov model to identify representative motifs without predefined motif libraries.
• 🧩 Segmented global alignment — Efficiently handles ultra-long or complex STRs while maintaining base-level precision.
• 🔍 Smart interval merging — Applies an interval-gain decision strategy to accurately resolve overlapping STRs.
• 🧬 Enhanced detection in complex regions — Identifies confounding or nested repeat regions (e.g., centromeric satellites) with a novel density-based concentration test.
• 💾 Lightweight & scalable — Requires few dependencies, easy to install and run, and supports multiple operating systems.

---

⚙️ Installation

Option 1: Install via pip

pip install faststr

Option 2: Install via conda

(coming soon)

conda install -c bioconda faststr

Option 3: Local installation (development)

git clone https://github.com/yourname/faststr.git
cd faststr
pip install -e .

---

🚀 Quick Start

Basic Command

faststr [--strict | --normal | --loose] [--default] genome.fa

Example

faststr --strict --default genome.fa

This runs FastSTR in strict mode using the default model to identify STRs in the genome.fa file.

---

⚡ Command Line Options

Argument	Type	Default	Description
match	int	2	Match score
mismatch	int	5	Mismatch score
gap_open	int	7	Gap opening penalty
gap_extend	int	3	Gap extension penalty
p_indel	int	15	Indel percentage threshold
p_match	int	80	Match percentage threshold
score	int	50	Alignment score threshold
quality_control	bool	False	Enable read-level quality control
DNA_file	str	—	Path to DNA FASTA input
-f	str	—	Output directory
-s	int	1	Start index
-e	int	0	End index
-l	int	15000	Sub-read length
-o	int	1000	Overlap length
-p	int	1	Number of CPU cores
-b	float	0.045	Motif coverage threshold

---

🧠 Alignment Modes

Mode	Description
--strict	High precision, recommended for curated assemblies
--normal	Balanced mode, suitable for most datasets
--loose	High sensitivity, tolerant of mismatches

---

🧬 Model Presets

Preset	Description
--default	Standard scoring model
(future) --sensitive	Optimized for noisy long reads
(future) --speed	Optimized for large-scale detection

---

📥 Input & Output

Input

• DNA sequences in FASTA format

Output

File Pattern	Description
*detail.dat	Contains all STR positions and motifs, quality statistics for each STR, and STR counts per chromosome.
*align.dat	Detailed alignment of all STRs against reference STRs, including mismatches and indels.
*.csv	Merged STR intervals with representative motifs and summary statistics for each interval.
*.log	Processing logs.

---

🧪 Usage

1️⃣ Identify STRs in a genome

faststr --normal --default human_genome.fa

2️⃣ Use multiple cores

faststr --strict --default genome.fa -p 8

3️⃣ Example run

faststr --strict --default NC_024512.1.fna -p 8

---

📈 Performance

Dataset	Genome Size	Tool	Runtime	Recall	Precision
Human (T2T)	2.94 G	TRF	18 h 31 min	-	-
		FastSTR	1 h 13 min	0.950	0.994
Mouse (GRCm39)	2.57 G	TRF	1 h 41 min	-	-
		FastSTR	38 min	0.966	0.997
Zebrafish (GRCz11)	1.58 G	TRF	2 h 51 min	-	-
		FastSTR	25 min	0.945	0.998

Note: TRF is used as the ground-truth. FastSTR runs based on 72 CPUs.

---

📚 Citation

If you use FastSTR in your research, please cite:

Xingyu Liao et al.,
Efficient Identification of Short Tandem Repeats via Context-Aware Motif Discovery and Ultra-Fast Sequence Alignment,
Nat. XXX, 2026.

---

📄 License

This project is licensed under the MIT License.
See LICENSE for more details.

---

🧾 Changelog

v1.0.0 (2025)

• Initial release of FastSTR
• Supports three alignment modes and one default model
• Implemented parallel computation
• Added .csv, .dat, .log outputs

---