ENSAS - Expression Neighborhood Sequence Association Study

Identifying associations of de novo noncoding variants with phenotypes through integration of gene expression, sequence and sex information

https://doi.org/10.1186/s13059-025-03619-1

Requirements

numpy==1.24.3
pandas==2.0.3
scikit_learn==1.3.0
scipy==1.12.0

Basic usage

usage: model.py [-h] [-v [VARIANTS]] [-g [GENES]] [-o [OUTPUT]]
                [-n [NEIGHBORS]] [-k [KMER]] [-c [GENE_COL]]

options:
  -h, --help            show this help message and exit
  -v [VARIANTS], --variants [VARIANTS]
                        Variants file
  -g [GENES], --genes [GENES]
                        Gene matrix file
  -o [OUTPUT], --output [OUTPUT]
                        Output path
  -n [NEIGHBORS], --neighbors [NEIGHBORS]
                        Size of each neighborhood, default 1000
  -k [KMER], --kmer [KMER]
                        Length of k-mers, default 6
  -c [GENE_COL], --gene_col [GENE_COL]
                        Column for gene assignment, default: GeneNAMEOuterTSS

The input variants file must have columns "Proband" (binary proband-sibling phenotype labels), "sequence" (DNA sequence), and a column specifying gene assignment for each variant (default: "GeneNAMEOuterTSS"). Sample files can be found in the data folder. The gene matrix file must be tab-delimited, with a header and an index column. A sample gene matrix file can be found in the data folder.

Output format

The columns in the output file are:

• Gene name
• The local GC content Mann-Whitney U effect size, computed from dividing the U-statistic by the total number of comparisons
• The local GC content Mann-Whitney U p-value
• The K-mer Naive Bayes model Mann-Whitney U effect size on the testing fold
• The K-mer Naive Bayes model Mann-Whitney U p-value on the testing fold
• The local GC content Mann-Whitney U effect size on the testing fold
• The local GC content Mann-Whitney U p-value on the testing fold All p-values reported are uncorrected.