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
|
# Permutations
Currently we use gemma-wrapper to compute the significance level - by shuffling the phenotype vector 1000x.
As this is a lengthy procedure we have not incorporated it into the GN web service. The new bulklmm may work
in certain cases (genotypes have to be complete, for one).
Because of many changes gemma-wrapper is not working for permutations. I have a few steps to take care of:
* [ ] read R/qtl2 format for phenotype
# R/qtl2 and GEMMA formats
See
=> data/R-qtl2-format-notes
# One-offs
## Phenotypes
For a study Dave handed me phenotype and covariate files for the BXD. Phenotypes look like:
```
Record ID,21526,21527,21528,21529,21530,21531,21532,21537,24398,24401,24402,24403,24404,24405,24406,24407,24408,24412,27513,27514,27515,27516,
27517
BXD1,18.5,161.5,6.5,1919.450806,3307.318848,0.8655,1.752,23.07,0.5,161.5,18.5,6.5,1919.450806,3307.318848,0.8655,1.752,0.5,32,1.5,1.75,2.25,1.
25,50
BXD100,x,x,x,x,x,x,x,x,x,x,x,x,x,x,x,x,x,x,x,x,x,x,x
BXD101,20.6,176.199997,4.4,2546.293945,4574.802734,1.729,3.245,25.172001,0.6,176.199997,20.6,4.4,2546.294189,4574.802734,1.7286,3.2446,0.6,32,
1.875,2.375,2.75,1.75,38
BXD102,18.785,159.582993,6.167,1745.671997,4241.505859,0.771,2.216,22.796667,0.25,159.583328,18.785,6.166667,1745.672485,4241.506348,0.770667,
2.216242,0.25,28.08333,1.5,2,2.875,1.5,28.5
...
```
which is close to the R/qtl2 format. GEMMA meanwile expects a tab delimited file where x=NA. You can pass in the column number with the -n switch. One thing GEMMA lacks it the first ID which has to align with the genotype file. The BIMBAM geno format, again, does not contain the IDs. See
=> http://www.xzlab.org/software/GEMMAmanual.pdf
What we need to do is create and use R/qtl2 format files because they can be error checked on IDs and convert those, again, to BIMBAM for use by GEMMA. In the past I wrote Python converters for gemma2lib:
=> https://github.com/genetics-statistics/gemma2lib
I kinda abandoned the project, but you can see a lot of functionality, e.g.
=> https://github.com/genetics-statistics/gemma2lib/blob/master/gemma2/format/bimbam.py
We also have bioruby-table as a generic command line tool
=> https://github.com/pjotrp/bioruby-table
which is an amazingly flexible tool and can probably do the same. I kinda abandoned that project too. You know, bioinformatics is a graveyard of projects :/
OK, let's try. The first step is to convert the phenotype file to something GEMMA can use. We have to make sure that the individuals align with the genotype file(!). So, because we work with GN's GEMMA files, the steps are:
* [X] Read the JSON layout file - 'sample_list' is essentially the header of the BIMBAM geno file
* [X] Use the R/qtl2-style phenotype file to write a correct GEMMA pheno file (multi column)
* [ ] Compare results with GN pheno output
Running GEMMA by hand it complained
```
## number of total individuals = 235
## number of analyzed individuals = 26
## number of covariates = 1
## number of phenotypes = 1
## number of total SNPs/var = 21056
## number of analyzed SNPs = 21056
Calculating Relatedness Matrix ...
rsm10000000001, X, Y, 1, 1, 0, 0, 0, 1, 1, 0, 1, 1, 0, 0, 1, 0, 0, 0, 0, 1, 1, 1, 0, 1, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 1, 0, 1, 1, 0, 1, 1, 1, 0, 0, 1, 0, 1, 1, 1, 0, 1, 1, 0, 0, 0, 0, 0, 1, 0, 1, 1, 0, 1, 1, 1, 1, 0, 0, 0, 1, 0, 1, 0, 0, 0, 1, 0, 1, 0, 0, 1, 1, 1, 1, 0, 1, 0, 0, 1, 1, 1, 1, 1, 0.5, 0, 1, 0, 1, 0.5, 0, 1, 0, 0, 0, 1, 1, 0, 0.5, 1, 1, 0.5, 0, 1, 1, 0, 0, 1, 1, 1, 0, 1, 1, 1, 0.5, 0, 0, 0, 0, 1, 1, 1, 1, 0, 0, 1, 0, 1, 0, 0, 0, 1, 0, 0, 0, 0, 0, 1, 0, 0, 1, 1, 1, 1, 1, 1, 0, 1, 0, 0, 1, 0, 1, 0, 1, 1, 0, 0, 1, 1, 0.5, 1, 1, 1, 0, 0, 1, 1, 1, 1, 0, 0, 0, 0.5, 0, 0, 0.5, 0, 1, 0, 1, 0, 0, 1, 0.5, 0, 1, 0, 0.5, 1, 1, 1, 1, 0.5, 0, 0, 0.5, 1, 0.5, 0.5, 0.5, 1, 0.5, 1, 0.5, 0.5, 0, 0, 0, 0.5, 1, 0.5, 0, 0, 0.5, 0, 0, 1, 0, 0.5, 1, 0.5, 0.5, 0.5, 1, 0.5, 0.5, 0.5
237 != 235
WARNING: Columns in geno file do not match # individuals in phenotypes
ERROR: Enforce failed for not enough genotype fields for marker in src/gemma_io.cpp at line 1470 in BimbamKin
```
GEMMA on production is fine. So, I counted BXDs. For comparison, GN's pheno outputs 241 BXDs. Daves pheno file has 241 BXDs (good). But when using my script we get 235 BXDs. Ah, apparently they are different from what we use on GN because GN does not use the parents and the F1s for GEMMA. So, my script should complain when a match is not made. Turns out the JSON file only contains 235 'mappable' BXDs and refers to BXD.8 which is from Apr 26, 2023. The header says `BXD_experimental_DGA_7_Dec_2021` and GN says WGS March 2022. So which one is it? I'll just go with latest, but genotype naming is problematic and the headers are not updated.
> MOTTO: Always complain when there are problems!
Luckily GEMMA complained, but the script should have also complained. The JSON file with 235 genometypes is not representing the actual 237 genometypes. We'll work on that in the next section.
Meanwhile let's add this code to gemma-wrapper. The code can be found here:
=> https://github.com/genetics-statistics/gemma-wrapper/blob/master/bin/rqtl2-pheno-to-gemma.py
## Genotypes
The pheno script now errors with
```
ERROR: sets differ {'BXD065xBXD102F1', 'C57BL/6J', 'DBA/2J', 'BXD077xBXD065F1', 'D2B6F1', 'B6D2F1'}
```
Since these are parents and F1s, and are all NAs in Dave's phenotypes, they are easy to remove. So, now we have 235 samples in the phenotype file and 237 genometypes in the genotype file (according to GEMMA). A quick check shows that BXD.geno has 236 genometypes. Same for the bimbam on production. We now have 3 values: 235, 236 and 237. Question is why these do not overlap.
### Genotype probabilities for GEMMA
Another problem on production is that we are not using the standard GEMMA values. So GEMMA complains with
```
WARNING: The maximum genotype value is not 2.0 - this is not the BIMBAM standard and will skew l_lme and effect sizes
```
This explains why we divide the effect size by 2 in the GN production code. Maybe it is a better idea to fix then geno files!
* [X] Generate BIMBAM file from GENO .geno files (via R/qtl2)
* [X] Check bimbam files on production
So we need to convert .geno files as they are the current source of genotypes in GN and contain the sample names that we need to align with pheno files. For this we'll output two files - one JSON file with metadata and sample names and the actual BIMBAM file GEMMA requires. I notice that I actually never had the need to parse a geno file! Zach wrote a tool `gn2/maintenance/convert_geno_to_bimbam.py` that also writes the GN JSON file and I'll take some ideas from that. We'll also need to convert to R/qtl2 as that is what Dave can use and then on to BIMBAM. So, let's add that code to gemma-wrapper again.
This is another tool at
=> https://github.com/genetics-statistics/gemma-wrapper/blob/master/bin/gn-geno-to-gemma.py
where the generated JSON file helps create the pheno file. We ended up with 237 genometypes/samples to match the genotype file and all of Dave's samples matched. Also, now I was able to run GEMMA successfully and passed in the pheno column number with
```
gemma -gk -g BXD-test.txt -p BXD_pheno_Dave-GEMMA.txt -n 5
gemma -lmm 9 -g BXD-test.txt -p BXD_pheno_Dave-GEMMA.txt -k output/result.cXX.txt -n 5
```
the pheno file can include the sample names as long as there are no spaces in them. For marker rs3718618 we get values -9 0 X Y 0.317 7.930689e+02 1.779940e+02 1.000000e+05 7.532662e-05. The last value translates to
```
-Math.log10(7.532662e-05) => 4.123051519468808
```
and that matches GN's run of GEMMA w.o. LOCO.
The next step is to make the -n switch run with LOCO on gemma-wrapper.
```
./bin/gemma-wrapper --loco --json -- -gk -g BXD-test.txt -p BXD_pheno_Dave-GEMMA.txt -n 5 -a BXD.8_snps.txt > K.json
./bin/gemma-wrapper --keep --force --json --loco --input K.json -- -lmm 9 -g BXD-test.txt -p BXD_pheno_Dave-GEMMA.txt -n 5 -a BXD.8_snps.txt > GWA.json
```
Checking the output we get
```
-Math.log10(3.191755e-05) => 4.495970452606926
```
and that matches Dave's output for LOCO and marker rs3718618. All good, so far. Next step permute.
## Permute
Now we have gemma-wrapper working we need to fix it to work with the latest type of files.
## Covariates
- [ ] Try covariates Dave
|
