PanelRecon

PanelRecon is a fast and highly sensitive tool to assign the most likely gene panel or exome to each fastq pair.

How it works

PanelRecon compares the k-mer diversity in each sample against a precomputed panel k-mer index. Top-scoring panels are identified based on the number of weighed supporting k-mers. Highly shared kmers across panels are penalized. For high-scoring panel candidates (≥95% similarity), a further refining step is performed relying on panel-specific k-mers.

PanelRecon ranks panels using an fscore-like:

• score = (1 + beta^2) * (panelCoverage * specificityPrecision) / (beta^2 * specificityPrecision + panelCoverage).

Where,

• panelCoverage = covered_panel_kmers / panel_unique_kmers,
• specificityPrecision: is the fraction of matched weighted k-mer evidence assigned to that panel (shared k-mers are penalized). A higher score means the panel is both broadly covered and more specific to the sample.

Installation

• To install zlib and htslib

sudo apt-get update
sudo apt-get install -y build-essential pkg-config zlib1g-dev libhts-dev

Requirements:

• C++17 compiler such as g++-
• make.
• htslib.
• zlib.

Build:

git clone https://github.com/GENCARDIO/PanelRecon.git
cd PanelRecon
make

Commands

PanelRecon has two subcommands:

1. index: build panel index files (.2bit) from BED panels and a reference FASTA.
2. find: scan FASTQ reads against those indexes and rank panels.

1. Index

./PanelRecon index \
  [--bed panel.bed | --bed_list beds.txt] \
  --fasta reference.fa \
  --output_dir /path/to/panel_index_dir \
  --kmer_size 31 \
  --minimizer_window 1

index minimizer notes:

• --minimizer_window controls k-mer subsampling while building indexes.
• Default is 1 (disabled, keep all k-mers).
• Values > 1 keep minimizers per window and reduce index size.

2. find

Scan FASTQ reads against panel indexes and write ranking output.

Example with a paired FASTQ:

./PanelRecon find \
  --index_dir /path/to/panel_index_dir \
  --fq1 sample_R1.fastq.gz \
  --fq2 sample_R2.fastq.gz \

Example with a list of fastqs --fastq_list:

./PanelRecon find \ 
  --index_dir /path/to/panel_index \
  --fastq_list fastq_list.txt \
  --output /path/to/output_ranks.tsv

fastq_list.txt must contain exactly one FASTQ file path per line.

Output

Panel ranking is written to --output (default: panel_ranks.tsv).

Example:

sample	scanned_reads	best_panel	best_score	score_margin_vs_next	best_panel_covered_kmers	best_panel_covered_kmers_pct
SAMPLE_A	100000	Exome.2bit	0.6731	0.0312	129884	67.31
SAMPLE_B	50000	GenePanel.2bit	0.4108	0.0457	54321	41.08