GWAS Simulation

comvert hmp to ped1, ped2, map file
SB1.ped, SB2.ped, SB.map

1, choose 20 markers for 30 times
(WD: /share/bioinfo/miaochenyong/GWAS/SB/20Markers-1To5Effect)
python ../choose_multi-markers.py SB.imputed.916.filtered.hmp 20 30 marker pheno

2, combine pheno, ped1, ped2 to intact ped file

python ../genCombine.py phenoPrefix 30 > combine.sh
parallel -j 30 < combine.sh

3, copy SB.map to 30 different SB-*.map
 python ../CPmapTOmore.py 30 SB-

4, *map, *ped to *bed, *bim, *fam
python ../generatePLINKcmd.py 30 SB- > PLINK.cmd
chmod 777 PLINK.cmd
parallel -j 6 < PLINK.cmd

5, run gemma
python ../generateGemmaCmd.py 30 SB- > gemma.cmd
chmod 777 gemma.cmd
parallel -j 6 < gemma.cmd

Calculate FDR value:
(WD: /share/bioinfo/miaochenyong/GWAS/SB/20Markers-1To5Effect-FDR)
1, shuffle pheno1.txt to 100 pheno*.txt
python ../shufflePheno.py pheno3.txt 100 pheno-shuffled

2, combine pheno, ped1, ped2 to intact ped file
python ../genCombine.py phenoPrefix 100 > combine.sh
parallel -j 100 < combine.sh

3, copy SB.map to 100 different SB-shuffle*.map
python ../CPmapTOmore.py  100 SB-shuffle-

4, *map, *ped to *bed, *bim, *fam
 python ../generatePLINKcmd.py 100 SB-shuffle- > PLINK.cmd
chmod 777 PLINK.cmd
parallel -j 10 < PLINK.cmd

5, run gemma
python ../generateGemmaCmd.py 100 SB-shuffle- > gemma.cmd
chmod 777 gemma.cmd
parallel -j 10 < gemma.cmd

6, calsulate FDR
cd output
python ../../calculateFDR.py SB-shuffle- 100 results.txt

Calculate average Power:
(WD: /share/bioinfo/miaochenyong/GWAS/SB/20Markers-1To5Effect/output)
python ../../calPower.py SB- marker 30 /share/bioinfo/miaochenyong/GWAS/SB/20Markers-1To5Effect-FDR/output/results.txt SB-
python ../../calAveragePower.py SB-

generage new effect 0.9+8
(WD: /share/bioinfo/miaochenyong/GWAS/SB/20Markers-0.9Effect)
ln -s /share/bioinfo/miaochenyong/GWAS/SB/20Markers-1To5Effect/markers-new* .
ln -s ../Imputed/SB.imputed.916.filtered.hmp .
python ../newEffect.py SB.imputed.916.filtered.hmp markers-new 30

事实证明：

平均数取8， 20， 100 模拟结果一样

effect value 取0.9 和0.9*20 结果也一样，

表面结果不同是由于FDR不同导致的。

观察average power in different MAF region:

WD: /share/bioinfo/miaochenyong/GWAS/SB/20Markers-0.9Effect20/output

python ../../DrawHist20Markers.py

WD: /share/bioinfo/miaochenyong/GWAS/SB/5Markers-0.9Effect100/output

$ python ../../DrawHist5Markers.py

可以看到随着MAF增大， power上升。从以上两图也可以推测出整体的MAF分布，多数markers都在0.01-0.1之间。

整体分布：

WD: /share/bioinfo/miaochenyong/GWAS/SB/Imputed

python ../DrawMAFHist.py SB.imputed.916.filtered.hmp

增加遗传率：

WD： /share/bioinfo/miaochenyong/GWAS/SB/5Markers-0.9Effect100

python ../genHeritability.py pheno9.txt 0.7 pheno9-0.7H.txt

上图是5个markers， 发现很多个体有相同的表型，对20个makers的进行作图：

一样的表型很少。

calculate average power of various heritability:

1,generate new phenotype data containing heritability

cd  /share/bioinfo/miaochenyong/GWAS/SB/5Markers-1To5Effect100

python ../genHeriPheno.py pheno 30 0.7 phenoH0.7-

cd /share/bioinfo/miaochenyong/GWAS/SB/5Markers-1To5Effect100-0.7H

mv /share/bioinfo/miaochenyong/GWAS/SB/5Markers-1To5Effect100/phenoH0.7-* .

cp /share/bioinfo/miaochenyong/GWAS/SB/5Markers-1To5Effect100/marker* .

python ../genCombine.py phenoPrefix 30 > combine.sh

parallel -j 30 < combine.sh

python ../CPmapTOmore.py 30 SB-

python ../generatePLINKcmd.py 30 SB- > PLINK.cmd
parallel -j 6 < PLINK.cmd

python ../generateGemmaCmd.py 30 SB- > gemma.cmd
parallel -j 6 < gemma.cmd

Statistical results in Sorghum:

统计结果图：

MAF distribution in Seteria Italic:

python DrawMAFHist.py Seteria.imputed.GT.txt

发现小于0.05的基本没有，应该是被过滤掉了。

去除SB和SI中MAF 小于0.05的markers!

Transfer SI GT format to HMP format(SI directory):

python  GT2HMP.py Seteria.imputed.GT.txt Seteria.imputed.hmp

SI 有726080 个markers

WD: SB_VS_SI/

python FilterMAF.py SB.imputed.916.filtered.hmp SB.filteredMAF.hmp SB剩余198629 markers

python FilterMAF.py Seteria.imputed.hmp Seteria.filteredMAF.hmp SI剩余725588 markers

重新画MAF分布图 看两者是否相近，相近的话随机选择marker!

SB MAF filtered:

SI MAF filtered:

select 198629 markers randomly from 725588 markers in SI:

python  selectMarkers.py SI.filteredMAF.hmp 198629 SI.filteredMAF198629.hmp

重新做分布图：

cmiao

UNL

beadle center