-
Notifications
You must be signed in to change notification settings - Fork 8
/
slides.Rmd
993 lines (672 loc) · 24 KB
/
slides.Rmd
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
---
title: "Analysis of Genetic Data 2: Mapping Genome-Wide Associations"
author: Peter Carbonetto
output:
beamer_presentation:
template: docs/beamer.tex
fig_crop: false
keep_tex: false
fig_caption: false
pandoc_args: "--highlight-style=pygments"
---
```{r knitr-options, echo=FALSE}
knitr::opts_chunk$set(comment = "#",collapse = TRUE,fig.align = "center",
results = "hide",fig.show = "hide",fig.keep = "none",
message = FALSE,warning = FALSE)
```
Workshop aims
=============
1. Implement basic steps of a genome-wide association analysis.
2. Understand how phenotype and genotype data for a genome-wide
association study (GWAS) are encoded in computer files.
3. Understand some of the benefits (and complications) of using
linear-mixed models (LMMs) for GWAS.
4. Use command-line tools to inspect and manipulate phenotype and
genotype data.
Workshop aims
=============
+ This is a *hands-on workshop*---you will get the most out of this
workshop if you work through the exercises on your computer.
+ The examples are intended to run either on the RCC cluster or on
a laptop with Linux or macOS (not Windows).
+ About 15 GB will be needed to store all the the data and output files.
+ *Note: the examples may not produce the exact same result on your
laptop. Using the RCC cluster is recommended.*
Software we will use today
==========================
1. GEMMA
2. R
3. Several R packages: data.table, ggplot2, cowplot and stringr.
4. Basic shell commands such as "cat" and "wc".
Our research task
=================
Identify and interpret genetic loci contributing to tibia bone
development in mice.
1. We will use data from this study:
- Nicod *et al* (2016). "Genome-wide association of multiple
complex traits in outbred mice by ultra-low-coverage
sequencing." *Nature Genetics* **48:** 912--918.
2. To map tibia development genes, we will assess support (*p*-values)
for statistical association between tibia length and 353,697 SNP
genotypes.
3. We will visualize the association signal to identify candidate
genes for tibia bone development.
Background on study
===================
+ Samples are from an outbred mouse population, "Carworth Farms White"
(CFW).
+ 200 measurements (blood concentrations, muscle weights, *etc*) were
taken for each mouse.
+ Genotypes were obtained from low-coverage DNA sequencing.
+ Size of data is comparable to human GWAS: 2,000 samples,
350,000 SNPs.
+ Linkage disequilibrium (LD) does not decay as rapidly as humans, so
mapping resolution will not be as good as a human GWAS.
+ Data are publicly available.
Outline of workshop
===================
+ **Preliminaries**
+ Programming challenges:
1. Setting up your environment for an association analysis.
2. Preparing the data for an association analysis.
3. Running association analyses in GEMMA.
4. Visualizing and interpreting the results of an association
analysis.
Preliminaries
=============
+ WiFi.
+ Power outlets.
+ Reading what I type.
+ Pace, questions (e.g., keyboard shortcuts).
+ YubiKeys.
+ What to do if you get stuck.
Preliminaries
=============
+ The workshop packet is a repository on GitHub. Go to:
- github.com/rcc-uchicago/genetic-data-analysis-2
+ Download the workshop packet to your computer.
+ If necessary, unzip the ZIP file.
What's included in the workshop packet
======================================
+ **slides.pdf:** These slides.
+ **slides.Rmd:** R Markdown source used to create these slides.
+ **download_genotypes.sh:** Script for downloading the genotype data.
+ **functions.R**: Some R functions for loading the data and creating
plots.
+ **format.genotypes.for.gemma.R:** Script to convert genotype data
into a format suitable for GEMMA.
+ **CFW_measures.txt:** a large table with 200 columns (phenotypes)
and 2,117 rows (samples).
+ **CFW_covariates.txt:** covariate data for 2,117 CFW mice.
+ **listof1934miceusedforanalysis.txt:** ids of the 1,934 mice
used in the analyses of Nicod *et al* (2016).
Outline of workshop
===================
+ Programming challenges:
1. **Setting up your environment for an association analysis.**
2. Preparing the data for an association analysis.
3. Running association analyses in GEMMA.
4. Visualizing and interpreting the results of an association
analysis.
Challenge #1: Setting up your environment for GWAS
==================================================
+ Aim: Configure your computing environment for the next programming
challenges.
+ Steps:
1. Connect to midway2.*
2. Clone git repository.*
3. Download genotype data.
4. Download GEMMA.
3. Connect to a midway2 compute node.*
4. Launch R, and check your R environment.
5. Set up R for plotting.
Connect to midway2
==================
+ **If you have an RCC account:** I'm assuming you already know how to
connect to midway2. *ThinLinc is recommended if you do not know how
activate X11 forwarding in SSH.*
+ See: rcc.uchicago.edu/docs/connecting
+ **If you do not have an RCC account:** I can provide you with a
Yubikey. This will give you guest access to the RCC cluster (see the
next slide).
Using the Yubikeys
==================
+ Prerequisites:
1. SSH client (for Windows, please use **MobaXterm**)
2. USB-A port
+ Steps:
1. Insert Yubikey into USB port.
2. Note your userid: `rccguestXXXX`, where `XXXX` is the
last four digits shown on Yubikey.
3. Follow instructions to connect to midway2 via SSH, replacing
the cnetid with your `rccguestXXXX` user name:
rcc.uchicago.edu/docs/connecting
4. When prompted for password, press lightly on metal disc.
+ Important notes:
- Yubikeys do not work with ThinLinc.
- *Please return the Yubikey at the end of the workshop.*
Clone git repository
====================
Clone the workshop packet in your home directory on midway2 (**note:**
there are no spaces in the URL below):
```{bash download-packet, eval=FALSE}
cd $HOME
git clone https://github.com/rcc-uchicago/
genetic-data-analysis-2.git
cd genetic-data-analysis-2
```
Alternatively, it may be simpler to download and unzip the file
containing the contents of the git repository:
```{bash download-packet-easier, eval=FALSE}
cd $HOME
wget -O repo.zip https://tinyurl.com/2p9caa2m
unzip repo.zip
mv genetic-data-analysis-2-master \
genetic-data-analysis-2
```
Download genotype data
======================
The genotype data are too large to fit in the git repository. So we
need to download these data separately.
+ Go to: http://mtweb.cs.ucl.ac.uk/dosages
+ Download files for chromosomes 1--19.
+ You may find it easier to run the provided bash script:
```{bash download-genotypes, eval=FALSE}
bash download_genotypes.sh
```
+ After following these steps, you should have a total of 19 files
with extension ".RData" in the "genetic-data-analysis-2" folder.
Alternatively, download the prepared data (faster):
```{bash download-genotypes-midway2, eval=FALSE}
wget users.rcc.uchicago.edu/~pcarbo/cfwgeno.tar.gz
tar zxvf cfwgeno.tar.gz
```
Download GEMMA
==============
Download GEMMA 0.96. Go to:
+ https://github.com/genetics-statistics/GEMMA/releases/tag/v0.96
On midway2, you can run these commands to download and install GEMMA
0.96 for Linux:
```{bash get-gemma, eval=FALSE}
wget http://bit.ly/2I67fBV -O gemma.linux.gz
gunzip gemma.linux.gz
mv gemma.linux gemma
chmod 700 gemma
./gemma -h
```
Connect to a midway2 compute node
=================================
Set up an interactive session on a midway2 compute node with 4 CPUs
and 10 GB of memory:
```{bash connect-midway2, eval=FALSE}
screen -S workshop
sinteractive --partition=broadwl \
--reservation=gda2_workshop --cpus-per-task=4 \
--mem=10G --time=3:00:00 --account=pi-bigcheese
echo $HOSTNAME
```
Launch R
========
Start up an interactive R session. On midway2, you can start up R with
these commands:
```{bash start-R, eval=FALSE}
module load R/4.2.0
which R
R --no-save
```
*Note: You can use R or RStudio.*
Check your R environment
========================
Check the version of R you are running:
```{r check-version}
sessionInfo()
```
Check that you are starting with an empty environment:
```{r check-environment}
ls()
```
Check that you have the correct working directory---it should be set
to the "genetic-data-analysis-2" repository:
```{r check-wd}
getwd()
```
Set up R for plotting
=====================
Make sure you can display graphics in your current R session.
```{r plot-cars}
library(ggplot2)
data(cars)
qplot(cars$dist,cars$speed)
```
You should see a scatterplot. If not, your connection is not set up to
display graphics. On midway2, an alternative is to save the plot to a
file, and download the file to your computer using sftp or SAMBA. See:
+ rcc.uchicago.edu/docs/data-transfer
Another alternative is to run R or RStudio on your computer, and set
your working directory to the genetic-data-analysis-2 directory
mounted via SAMBA.
Quit R
======
We will quit R, and return to it later.
```{r quit, eval=FALSE}
quit()
```
Outline of workshop
===================
+ Preliminaries
+ Programming challenges:
1. Setting up your environment for an association analysis.
2. **Preparing the data for an association analysis.**
3. Running association analyses in GEMMA.
4. Visualizing and interpreting the results of an association
analysis.
Challenge #2: Prepare the data for GWAS
=======================================
+ Aim: Perform an exploratory analysis of the phenotype and covariate
data, and prepare the data for association analysis with GEMMA.
+ Steps:
1. Prepare genotype data for GEMMA.
2. Import phenotype and covariate data into R.
3. Explore phenotype and covariate data in R.
4. Save phenotype and covariate data for GEMMA.
Prepare the genotype data for GEMMA
===================================
Processing the genotype data is more complicated, so I have provided
an R script to create, for each chromosome, a SNP annotation file and
a genotype file in the format used by GEMMA. Run the script in R:
```{bash convert-genotypes, eval=FALSE}
Rscript format.genotypes.for.gemma.R
```
This will take a few minutes to run.
+ Note that you can skip this step if you downloaded the
`cfwgeno.tar.gz` file.
Prepare the genotype data for GEMMA
===================================
The genotype data need to be combined into one file. This can be
easily done with the `cat` command, which combines by lines (rows):
```{bash merge-genotype-data, eval=FALSE}
cat chr*.geno.txt > geno.txt
```
We do the same for the SNP annotations:
```{bash merge-snp-data, eval=FALSE}
cat chr*.map.txt > map.txt
```
+ Again, you can skip this step if you downloaded the
`cfwgeno.tar.gz` file.
Import phenotypes and covariates into R
=======================================
Start up R. On midway2, run:
```{bash start-R-2, eval=FALSE}
R --no-save
```
Before continuing, verify that your R working directory is set to the
"genetic-data-analysis-2" repository:
```{r check-wd-2}
getwd()
```
Import phenotypes and covariates into R
=======================================
Load the phenotype and covariate data:
```{r load-pheno-data}
source("functions.R")
dat1 <- read.pheno("CFW_measures.txt")
dat2 <- read.pheno("CFW_covariates.txt")
```
Merge the two data frames, dropping the redundant "id" column:
```{r merge-pheno-data}
pheno <- cbind(dat1,dat2[-1])
```
Select samples
==============
We will use the same 1,934 samples that were used in the published
association analysis (Nicod *et al*, 2016). First, load the selected
sample ids:
```{r load-ids}
ids <-
read.table("listof1934miceusedforanalysis.txt",
stringsAsFactors = FALSE)
ids <- ids[[1]]
```
Next, select the rows of the phenotype table that match up with the
ids:
```{r select-ids}
rows <- match(ids,pheno$Animal_ID)
pheno <- pheno[rows,]
```
Examine tibia length
====================
Take a closer look at the data for the phenotype of interest, tibia
length (units are "mm"):
```{r summarize-tibia-length}
summary(pheno$Tibia.Length)
```
Plot the distribution of tibia length:
```{r plot-tibia-length-dist}
library(ggplot2)
library(cowplot)
theme_set(theme_cowplot())
p1 <- ggplot(pheno,aes(Tibia.Length)) +
geom_histogram(fill = "darkblue",
na.rm = TRUE)
print(p1)
```
Examine tibia length vs. body weight
====================================
The study authors recommend using body weight as a covariate in the
association analysis (units are "g"). What do the data tell us about
the relationship between the two?
```{r summarize-body-weight}
summary(pheno$Weight.Diss)
```
Visualize the relationship between tibia length and body weight:
```{r plot-tibia-vs-weight}
p2 <- qplot(pheno$Weight.Diss,
pheno$Tibia.Length,
na.rm = TRUE)
print(p2)
```
Examine tibia length vs. sex
============================
The study authors also recommend using sex as a covariate in the
association analysis. Let's examine the data:
```{r summarize-sex}
pheno <- transform(pheno,Sex = factor(Sex))
summary(pheno$Sex)
```
A boxplot is a common way to visualize the relationship between a
continuous variable (tibia length) and a categorical variable (sex):
```{r plot-tibia-vs-sex}
p3 <- ggplot(pheno,aes(Sex,Tibia.Length)) +
geom_boxplot(na.rm = TRUE)
print(p3)
```
Verify phenotype residuals
==========================
Compute the phenotypes after removing the linear effects of sex and
body weight:
```{r compute-residuals}
fit <- lm(Tibia.Length ~ Sex + Weight.Diss,pheno)
r <- resid(fit)
```
Plot the distribution of the residuals:
```{r plot-residuals}
p4 <- ggplot(data.frame(resid = r),aes(resid)) +
geom_histogram(fill = "darkblue")
print(p4)
```
Save phenotype data for GEMMA
=============================
Now that we have carefully examined the phenotype data, we are ready
to save the data in a format usable by GEMMA. First write the tibia
measurements to a new file, "pheno.txt":
```{r write-phenotype-data}
write.table(pheno$Tibia.Length,"pheno.txt",
row.names = FALSE,
col.names = FALSE)
```
Save covariate data for GEMMA
=============================
Next, write the covariates (body weight and sex) to new file,
"covar.txt", in the format used by GEMMA. *Sex needs to be encoded as
a number:*
```{r write-covariate-data}
sex <- as.numeric(pheno$Sex) - 1
dat <- data.frame(1,sex,pheno$Weight.Diss)
write.table(dat,"covar.txt",sep = " ",
row.names = FALSE,
col.names = FALSE)
```
Preparing the phenotype data: take-home points
==============================================
Most analyses will involve:
+ Carefully checking phenotype and covariate data for possible
issues, and resolving these issues (e.g., filtering out problematic
samples, transforming phenotypes).
+ Reformatting the data.
*Here we performed these steps in R.*
Preparing the genotype data: take-home points
=============================================
+ Typically, this is much more work (fortunately, the authors provided
the processed genotype data).
+ A common mistake is to save the phenotype and genotype samples in
different orders, which will lead to an incorrect association
analysis!
Outline of workshop
===================
+ Preliminaries
+ Programming challenges:
1. Setting up your environment for an association analysis.
2. Preparing the data for an association analysis.
3. **Running association analyses in GEMMA.**
4. Visualizing and interpreting the results of an association
analysis.
Challenge #3: Run association analyses
======================================
+ Aim: Compute *p*-values to assess support for association between
tibia length and SNPs on chromosomes 1--19.
+ Steps:
1. Run a basic association analysis in GEMMA.
2. Run an LMM-based association analysis in GEMMA.
3. Compare the two association analyses.
Basic association analysis in GEMMA
===================================
Now that we have installed GEMMA, and we have prepared the phenotype
and genotype data in the formats accepted by GEMMA, we are now ready
to run our first association analysis.
```{bash gemma-lm, eval=FALSE}
./gemma -p pheno.txt -c covar.txt \
-a map.txt -g geno.txt -notsnp \
-lm 2 -outdir . -o tibia
```
This should take no more than a few minutes to complete.
To view the results of the association alaysis, run
```{bash gemma-lm-view, eval=FALSE}
column -t -o ' ' tibia.assoc.txt | less -S
```
Basic association analysis in GEMMA
===================================
This GEMMA call creates two files:
+ **tibia.log.txt:** A summary of the association analysis.
+ **tibia.assoc.txt:** The full table of association results
(*p*-values, effect size estimates, *etc*).
LMM-based association analysis in GEMMA
=======================================
First, compute a kinship ("realized relatedness") matrix from the
genotypes.
```{bash gemma-kinship, eval=FALSE}
./gemma -p pheno.txt -g geno.txt \
-a map.txt -gk 1 -outdir . -o tibia
```
Next, compute LMM-based association statistics for each SNP:
```{bash gemma-lmm, eval=FALSE}
./gemma -p pheno.txt -c covar.txt \
-a map.txt -g geno.txt -notsnp \
-k tibia.cXX.txt -lmm 2 -outdir . \
-o tibia_lmm
```
LMM-based association analysis in GEMMA
=======================================
Since the LMM-based association analysis may take a while to complete
(perhaps an hour or more), I have provided files containing the
outputs of these two GEMMA commands:
+ **tibia.cXX.txt.gz:** The 1,934 x 1,934 kinship matrix.
+ **tibia_lmm.log.txt:** A summary of the association analysis.
+ **tibia_lmm.assoc.txt.gz:** The full table of association results.
To use these files, you need to uncompress them first:
```{bash gemma-lmm-gunzip, eval=FALSE}
gunzip tibia.assoc.txt.gz
gunzip tibia_lmm.assoc.txt.gz
```
Compare basic and LMM association analyses
==========================================
The LMM is expected to reduce inflation of small *p*-values; a high
level of inflation could indicate many false positive associations.
+ The q-q plot is commonly used to assess inflation. This test is
useful as a simple, heuristic diagnostic.
Launch R again, and import the results of the first GEMMA association
analysis:
```{r load-gemma-pvalues}
gwscan1 <- read.table("tibia.assoc.txt",
as.is = "rs",header = TRUE)
```
Compare basic and LMM association analyses
==========================================
Plot the observed *p*-values against the expected *p*-values under the
null distribution:
```{r plot-inflation-basic}
library(ggplot2)
library(cowplot)
theme_set(theme_cowplot())
source("functions.R")
p1 <- plot.inflation(gwscan1$p_lrt)
print(p1)
```
Next, we will compare this plot against the LMM q-q plot.
Compare basic and LMM association analyses
==========================================
Follow the same steps as before to load the GEMMA results and create a
genomic inflation plot:
```{r plot-inflation-lmm}
gwscan2 <- read.table("tibia_lmm.assoc.txt",
as.is = "rs",header = TRUE)
p2 <- plot.inflation(gwscan2$p_lrt)
print(p2)
```
Compare against the previous inflation plot.
LMM-based association analysis: take-home points
================================================
+ LMMs are now one of the most widely used approaches to reduce
inflation of false positive associations due to population structure
and hidden relatedness.
+ LMMs can be computationally intensive, especially for data
sets with many samples.
+ LMMs can *overcorrect* inflation. One remedy is the
"leave-one-chromosome-out" (LOCO) approach.
Outline of workshop
===================
+ Preliminaries
+ Programming challenges:
1. Setting up your environment for an association analysis.
2. Preparing the data for an association analysis.
3. Running association analyses in GEMMA.
4. **Visualizing and interpreting the results of an association
analysis.**
Challenge #4: Visualize and interpret results of association analysis
=====================================================================
+ Aim: Create plots to gain insight into the results of the
association analysis.
+ Steps:
1. Summarize "genome-wide" association signal for tibia length.
2. Visualize strongest association signal.
3. Based on the association signal, identify candidate genes.
4. Visualize the relationship between tibia length and the
top SNP association.
Plot genome-wide scan for tibia length
======================================
Using GEMMA, we computed association *p*-values at 353,697 SNPs on
chromosomes 1--19. Now let's plot these *p*-values to get a visual
summary of the association results. First, load the results of the LMM
association analysis:
```{r load-gemma-lmm-pvalues}
gwscan <- read.table("tibia_lmm.assoc.txt",
as.is = "rs",header = TRUE)
```
Plot genome-wide scan for tibia length
======================================
I wrote a function that uses `ggplot2` to create a "Manhattan plot"
showing the results of the association analysis.
```{r plot-gwscan}
library(ggplot2)
library(cowplot)
theme_set(theme_cowplot())
source("functions.R")
p1 <- plot.gwscan(gwscan)
print(p1)
```
This gives us a "genome-wide" view of the association signal.
Examine the association signal on chromosome 6
==============================================
Here we look more closely at the strongest association signal on
chromosome 6. First, create a data frame with the results for
chromosome 6 only:
```{r get-gwscan-chr6}
gwscan_chr6 <- subset(gwscan,chr == 6)
```
Next, plot the *p*-values against the base-pair position of the SNPs.
```{r plot-pvalues-chr6}
p2 <- plot.region.pvalues(gwscan_chr6)
print(p2)
```
Examine the association signal on chromosome 6
==============================================
It appears that the strongest association signal is isolated to
positions of 145 Mb and greater. Let's zoom in on that region:
```{r plot-pvalues-chr6-locus}
dat <- subset(gwscan_chr6,ps > 145e6)
p3 <- plot.region.pvalues(dat)
print(p3)
```
Let's find the SNP on chromosome 6 that is most strongly
associated with tibia length:
```{r top-marker}
i <- which.min(gwscan_chr6$p_lrt)
gwscan_chr6[i,]
```
Visualize genotype-phenotype relationship
=========================================
Above, we identified the SNP that is most strongly associated with
tibia length.
+ Next, we examine the relationship between tibia length
and the genotype at this SNP.
Load the GEMMA phenotype data:
```{r load-pheno-gemma}
pheno <- read.table("pheno.txt")
pheno <- pheno[[1]]
```
Visualize genotype-phenotype relationship
=========================================
Load the GEMMA genotype data for chromosome 6, and retrieve the
genotypes for the top SNP. To obtain discrete genotype values (0, 1,
2), round the numeric values to the nearest integer.
```{r load-geno-gemma}
command <- paste("grep snp-6-146795232 geno.txt >",
"geno_chr6_146795232.txt")
system(command)
geno <- read.table("geno_chr6_146795232.txt",
sep = " ")
geno <- as.numeric(geno)
geno <- geno[-(1:3)]
```
Visualize genotype-phenotype relationship
=========================================
If we convert the genotype to a "factor", we can visualize the
relationship with a boxplot.
```{r geno-num2factor}
x <- round(geno)
x <- factor(x,0:2)
levels(x) <- c("GG","AG","AA")
```
A is the alternative allele, so a 2 corresponds to genotype "AA".
```{r plot-pheno-vs-geno}
pdat <- data.frame(pheno = pheno,x = x)
p <- ggplot(pdat,aes(x,pheno)) +
geom_boxplot(na.rm = TRUE)
print(p)
```
This provides a visual summary of the genotype-phenotype relationship.
We can also quantify the relationship:
```{r lm-pheno-vs-geno}
fit <- lm(y ~ x,data.frame(x = geno,
y = pheno))
summary(fit)
```
Recap
=====
+ We implemented the basic steps of a GWAS using GEMMA, R and
command-line tools such as "cat".
+ We saw that much of the effort went into data processing and quality
control.
+ We did not discuss how to determine an appropriate significance
threshold. This is a complicated question. *My advice:* first take
an exploratory approach to identify the strongest association
signals, and determine significance later.