AutoFFinder

AutoFFinder is a runtime-reconfigurable hardware-software co-design for multi-gRNA CRISPR/Cas9 off-target search:

• ReLev (FPGA stage): high-throughput Levenshtein automata candidate generation.
• PostAutoFFinder (CPU stage): exact alignment reconstruction and biological filtering.

This repository stores the project code and is organized as a two-stage pipeline. It is not a single one-command tool.

Repository structure

• ReLev/ – FPGA code and host code for candidate generation (submodule).
• PostAutoFFinder/ – Java post-processing code that consumes ReLev outputs and produces final off-target CSV results.

Stage 1: Run ReLev separately

Use the ReLev documentation for build and execution details:

• ReLev README

If cloning from scratch, make sure the submodule is available:

git submodule update --init --recursive

Stage 2: Run PostAutoFFinder

Requirements

• Java 17 (tested with OpenJDK 17)

Compile

javac -d bin PostAutoFFinder/*.java

Run

java -cp bin PostAutoFFinder.AutoOffTargetSearchAlign \
	<Genome reference FASTA> \
	<gRNA file path> \
	<Output prefix> \
	<maxE> <maxM> <maxMB> <maxB> \
	<Threads> \
	<Best-in-window> <Best-window-size> \
	<PAM> <Allow PAM edits> \
	<ReLev output folder>

AutoOffTargetSearchAlign expects 13 positional arguments in exactly this order.

Input expectations and formats

1) Genome reference FASTA

• Multi-FASTA is supported.
• Each record is split into per-chromosome text files internally by PostAutoFFinder.

2) gRNA file (<gRNA file path>)

• Plain text file, one guide per line.
• Each line should include the guide with PAM suffix (for example, NNNNNNNNNNNNNNNNNNNNNGG).
• Allowed characters are DNA bases and N (see sgRNAs.txt for an example).

3) ReLev output folder (<ReLev output folder>)

This folder must contain ReLev candidate matches per chromosome and strand in text files named:

• <chromosome>_fw.txt for forward strand
• <chromosome>_rc.txt for reverse-complement strand

where <chromosome> must match the chromosome filenames generated from the FASTA headers.

Each file must contain one candidate match per line in this format:

<end_position>:<target_id>

• <end_position>: integer genomic end position reported by ReLev.
• <target_id>: zero-based index of the gRNA in the input gRNA file.

Arguments

1. Genome reference FASTA: Path to the FASTA genome.
2. gRNA file path: Path to text file containing guides (one per line, including PAM).
3. Output prefix: Output prefix. Final file is written as <Output prefix>.csv.
4. maxE: Maximum total edits (integer).
5. maxM: Maximum mismatches when no bulges are used (integer).
6. maxMB: Maximum mismatches when bulges are used (integer).
7. maxB: Maximum bulges (integer).
8. Threads: Number of CPU threads for post-processing (integer).
9. Best-in-window: true or false.
10. Best-window-size: Window size used when argument 9 is true (integer).
11. PAM: PAM sequence, e.g. NGG.
12. Allow PAM edits: true or false.
13. ReLev output folder: Folder containing ReLev candidate files (*_fw.txt, *_rc.txt).

Output format

PostAutoFFinder writes one CSV file (<Output prefix>.csv) with columns:

• Chromosome
• Strand
• EndPosition
• Target
• SiteSeqPlusMaxEditsBefore
• #Edit
• AlignedTarget
• AlignedText
• #Mismatches
• #Bulges