ClusterAnalyzer

ClusterAnalyzer is a clustering algorithm for prioritizing likely true signals from an excess of false positives. The tool is intended for genome-wide association studies with minimal sample size which suffer from insufficient power. Users are reminded to interpret the results with caution.

Prerequisite

• Python >= v3.9
• NumPy >= v1.23.4
• Pandas >= v1.5.1

Usage

usage: clusterAnalyzer.py [-h] --assoc ASSOC [--bfile BFILE] [--ld LDFILE] [--min_rows MIN_ROWS] [--mae MAE] [--outdir OUTDIR]
                          [--verbose VERBOSE]

options:
  -h, --help           show this help message and exit
  --assoc ASSOC        summary statistics from PLINK (*.logistic)
  --bfile BFILE        PLINK binary file prefix
  --ld FILE            LD statistics file
  --min_rows MIN_ROWS  Minimum SNP# requirement for a signal to be assessed
  --mae MAE            Mean absolute error threshold for determining LD-p value consistency
  --outdir OUTDIR      output directory (default: directory of summary statistics
  --verbose VERBOSE    level of verbosity (default: INFO)

Input/Output

ClusterAnalyzer takes association summary statistics as input and prioritizes signals that are likely to be true. Prioritized results are written to OUTDIR with variant_summary.assoc containing individual SNPs and cluster_sumamry.txt containing all prioritized genomic regions.

Chr	Start	End	SignalCount	Pmin	Pmax
1	163188121	163200062	15/9	0.0003528/0.03249	0.9963/0.9837
1	163241577	163284768	78/99	0.0005863/0.03234	0.9819/0.9716

• Chr: the chromosome where the signal is found
• Start, End: start and end coordinates of the signal
• SignalConut: number of SNPs with OR > 1 and number of SNPs with OR < 1 in the cluster
• Pmin, Pmax: minimum and maximum p values of SNPs with OR > 1 and OR < 1 in each cluster

Code correspondence

Louis SHE ([email protected])