cGMAS

cGMAS (Concordance-based GMAS) is a method for predicting functional SNPs for GMAS (genetically-modulated alternative splicing) events. GMAS events have associated SNPs that can serve as tag SNPs. If a SNP is functional for GMAS, we expect to see a concordant SNP genotype and alternative splicing pattern accross a large number of individuals. We can quanitify the concordance between genotype and splicing pattern using a concordance score (Si). This method cannot distinguish true functional and neutal SNPs if they are in perfect LD. Also, the method requires a large number of individuals.

Table of contents

A. Preprocessing

• Identify GMAS events (GMAS SNPs and GMAS exons)
• Determine genotype of candidate functional SNPs

B. Calculating concordance score (Si)

C. Filtering for candidate functional SNPs for GMAS

• Identify Si score peak
• Filter by mean Si scores and Si peak magnitude
• FDR corrections

D. Credits

A. Preprocessing

We can use the ASARP pipeline to get high-quality tag SNPs and GMAS events as input for this pipeline.

Get geontypes per tissue

Get the genotype of candidate SNPs. Possible sources: genotype information and RNA-seq data.

usage: gt.per.tissue.py [-h] -i infdir -a asasf -r refdir -t total -m mono -u
                        upperAR -l lowerAR -o outf

Get genotypes per tissue

optional arguments:
  -h, --help  show this help message and exit
  -i infdir   Input SNV count files dir
  -a asasf    Input asas file,asas events that are in at least X samples, sep by comma
  -r refdir   ref GT files directory
  -t total    total coverage to tell homozygous dbSNPs
  -m mono     mono coverage to tell homozygous dbSNPs
  -u upperAR  upper-bound allelic ratio for heterozygous dbSNPs (<=)
  -l lowerAR  lower-bound allelic ratio for heterozygous dbSNPs (>=)
  -o outf     Output file

B. Calculating concordance score (Si)

1. Get candidate functional SNPs

Union of GMAS events from all samples in a study. It does NOT matter which individuals we identified the GMAS events from.

usage: candid.bed.py [-h] -i ind -d indir -s suff -m min -o outf

Get candidate functional SNPs

optional arguments:
 -h, --help  show this help message and exit
 -i ind      Individual of interest
 -d indir    Input file directory
 -s suff     file suffix
 -m min      min number of tissues to decide heterozygous
 -o outf     Output file

2. Get genotypes of candidate functional SNPs

Each individual should have his or her specific list of SNPs and the genotype.

usage: tag.bed.py [-h] -i inf -a asasf -r ref -c cov -o outf

Get candidate functional SNP genotypes

optional arguments:
  -h, --help  show this help message and exit
  -i inf      Input candidate SNP bed file
  -a asasf    Input file with GMAS events that are in at least X samples
  -r ref      alleles count ref file
  -c cov      tag snv coverage
  -o outf     Output file

3. Calculate concordance scores

usage: splicing.concordance.py [-h] -i inf -d indiv -t tag -m maxD -o outf -a
                               anno -s search

Calculate concordance scores

optional arguments:
  -h, --help  show this help message and exit
  -i inf      Input candidate functional SNP bed file
  -d indiv    Input individual ID
  -t tag      tag snp bed file
  -m maxD     max d for RNA-seq defined tag snvs
  -o outf     Output file
  -a anno     gene annotation bed file
  -s search   max dist in nt from candidate functional snp to the AS exon to be
              tested; input "INF" to test all possible snp pairs within the same
              gene

C. Filtering for candidate functional SNPs for GMAS

1. Model Si scores of an event using GMM to get the Si score peak locations and to remove cases with shallow Si score peak magnitude.

usage: peak.si.rm.bg.py [-h] -i inf [-r ref] [-m min] -n N -o outf -b bin -p
                        het

Remove cases with shallow Si score peak magnitude

optional arguments:
  -h, --help  show this help message and exit
  -i inf      Input file prefix
  -r ref      SNVs to be filtered out
  -m min      min data points (individuals) in the Si distri
  -n N        Number of GMM components fitted
  -o outf     Output file
  -b bin      bin for randomization
  -p het      Percentage of heterozygous individuals

2. Get candidate functional SNPs based on Si, majorR, ... thresholds

• v1
  • number of individuals (n): 40
  • P-value testing whether the GMM is significantly different from Si = 1 (p): 0.1
  • Min Si (s): 0.8
  • Min % individuals in major GMM (m): 0.9 -> applies to only tag SNVs that have high enough (n)
  • Take out cases with all individuals who are homozygous (-M yes)

usage: get.causal.v1.py [-h] -i annoI -e annoE -r causalf -o outf -t tissue -s
                        si -p pval -n minPt -m major

Get functional SNPs - v1

optional arguments:
  -h, --help  show this help message and exit
  -i annoI    intron anno bed
  -e annoE    exon anno bed
  -r functionalf  ref functional si file dir
  -o outf     Output file
  -t tissue   tissue of interest
  -s si       min Si
  -p pval     min pval; pval is testing whether si is diff from 1
  -n minPt    min data points (indiv) per functional-exon-tag pair
  -m major    min membership ratio of the major component

• v2
  • number of individuals (n): 40
  • P-value testing whether the GMM is significantly different from Si = 1; Si = 0 (p): 0.1,0.01 -> >= 0.1 for Si = 1 and <= 0.01 for Si = 0

usage: get.causal.v2.py [-h] -i annoI -e annoE -r causalf -o outf -t tissue -p
                       pval -n minPt

Get functional SNPs - v2

optional arguments:
 -h, --help  show this help message and exit
 -i annoI    intron anno bed
 -e annoE    exon anno bed
 -r functionalf  ref functional si file dir
 -o outf     Output file
 -t tissue   tissue of interest
 -p pval     min pval; pval is testing whether si is diff from 1
 -n minPt    min data points (indiv) per functional-exon-tag pair

• v2b
  • number of individuals (n): 40
  • P-value testing whether the GMM is significantly different from Si = 1; Si = 0 (p): 0.05,0.05 -> <= 0.05 for Si = 1 and <= 0.05 for Si = 0
  • Min % individuals in major GMM (m): 0.9 -> applies to only tag SNVs that have high enough (n)
  • Min % het individuals (s; gtr): 0.95

usage: get.causal.v2b.py [-h] -i annoI -e annoE -r causalf -o outf -t tissue
                        -s gtr -p pval -n minPt -m major

Get functional SNPs - v2b

optional arguments:
 -h, --help  show this help message and exit
 -i annoI    intron anno bed
 -e annoE    exon anno bed
 -r functionalf  ref functional si file dir
 -o outf     Output file
 -t tissue   tissue of interest
 -s gtr      min GT ratio: RV/totalIndiv
 -p pval     min pval; pval is testing whether si is diff from 1
 -n minPt    min data points (indiv) per functional-exon-tag pair
 -m major    min membership ratio of the major component

3. Filter out non-functional SNPs

• v1

usage: detect.non-causal.py [-h] -i inf [-r ref] -s suff -o outf

Detect non-functional SNPs - v1

optional arguments:
 -h, --help  show this help message and exit
 -i inf      Input file prefix
 -r ref      ref file with SNV tested
 -s suff     suff
 -o outf     Output file

• v2

usage: detect.non-causal.v2.py [-h] -i inf [-r ref] -s suff -o outf

Detect non-functional SNPs - v2

optional arguments:
 -h, --help  show this help message and exit
 -i inf      Input file prefix
 -r ref      ref file with SNV tested
 -s suff     suffix
 -o outf     Output file

• v2b

usage: detect.non-causal.v2b.py [-h] -i inf [-r ref] -s suff -o outf

Detect non-functional SNPs - v2b

optional arguments:
 -h, --help  show this help message and exit
 -i inf      Input file prefix
 -r ref      ref file with SNV tested
 -s suff     suffix
 -o outf     Output file

4. FDR corrections

• v1 : fisherP.adjust.R (fisherP.adjust.sh)
• v2 : p.adjust.R (p0.adjust.sh)
• v2b: p.adjust.R (p0.adjust.v2b.sh)

D. Credits

• Esther Hsiao
• Grace Xiao lab