README.md 8.47 KB
Newer Older
eboulanger's avatar
eboulanger committed
1
2
3
# seaConnect--dataPrep

Scripts to prepare RADSeq data for analysis. 
eboulanger's avatar
eboulanger committed
4
- filtering steps for dDocent SNP output
eboulanger's avatar
eboulanger committed
5
- outlier detection
eboulanger's avatar
eboulanger committed
6
- file conversions, subsetting and renaming 
eboulanger's avatar
eboulanger committed
7

eboulanger's avatar
eboulanger committed
8
## 01-SNPfiltering 
eboulanger's avatar
eboulanger committed
9
10
11
12
13
14
15
16
17
18

script adapted from [ddocent tutorial](https://www.ddocent.com/filtering/) and additions

### run the script
move to the correct directory and make the output directories
```
cd ~/Documents/project_SEACONNECT/seaConnect--dataPrep/01-SNPfilters/
mkdir 01-Diplodus
mkdir 02-Mullus
```
eboulanger's avatar
eboulanger committed
19
20
21
22
set the species-specific arguments for the script to run on  
  $1 = input file (specify path if not in current directory)  
  $2 = output folder  
  $3 = species code  
eboulanger's avatar
eboulanger committed
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
```
bash filtering.sh ../00-rawData/01-Diplodus/sar_ddocent.vcf 01-Diplodus dip
bash filtering.sh ../00-rawData/02-Mullus/mullus.vcf 02-Mullus mul
```

### SNP filtering results for Mullus surmuletus					
				
| filtering step | filter for                                        | individuals retained | SNPs retained | run time (sec) | output |
| -------------- | ------------------------------------------------- | :------------------: | :-----------: | :------------: | ------ |
| step 0         | ddocent output data                               | 431                  | 49027         |                | mullus.vcf
| step 1         | call below 50%, mac < 3, min quality score = 30   | 431                  | 49027         | 71.00          | mullus.g5mac3.recode.vcf
| step 2         | min mean depth genotype call = 3                  | 431                  | 49027         | 74.00          | mullus.g5mac3dp3.recode.vcf
| step 3         | individuals > 50% missing data                    | 424                  | 49027         | 70.00          | mullus.g5mac3dplm.recode.vcf
| step 4         | sites > 5% missing data, maf 0.05, min meanDP = 5 | 424                  | 18727         | 14.00          | DP3g95maf05.recode.vcf
| step 5         | filter for allele balance                         |                      | 17965         |                | DP3g95maf05.fil1.vcf
| step 6         | filter out sites with reads from both strands     |                      | **SKIP**      |	             | 
| step 7         | ration mapping qualities ref vs alternate alleles |                      | 17546         |                | DP3g95maf05.fil3.vcf
| step 8         | paired status                                     |                      | 17546         |                | DP3g95maf05.fil4.vcf
| step 9         | remove sites quality score < 1/4 depth            |                      | 17546         |                | DP3g95maf05.fil5.vcf
| step 10        | depth x quality score cutoff	                     | 424                  | 15466         |	             | 
| step 11        | He > 0.6 & Fis > 0.5 & Fix < -0.5                 | 424                  | 15232         | 25 min         | DP3g95maf05.FIL.HFis.recode.vcf
| step 12        | rename                                            |                      |               |                | mul_all_filtered.vcf

### SNP filtering results for Diplodus sargus								
				
| filtering step | filter for                                        | individuals retained | SNPs retained | run time (sec) | output |
| -------------- | ------------------------------------------------- | :------------------: | :-----------: | :------------: | ------ |
| step 0         | ddocent output data                               | 297                  | 13362         |                | sar_ddocent_.vcf
| step 1         | call below 50%, mac < 3, min quality score = 30   | 297                  | 13362         | 14.00          | g5mac3.recode.vcf
| step 2         | min mean depth genotype call = 3                  | 297                  | 13362         | 14.00          | g5mac3dp3.recode.vcf
| step 3         | individuals > 50% missing data                    | 297                  | 13362         | 16.00          | g5mac3dplm.recode.vcf
| step 4         | sites > 5% missing data, maf 0.05, min meanDP = 5 | 297                  | 10389         | 11.00          | DP3g95maf05.recode.vcf
| step 5         | filter for allele balance                         | 297                  | 10202         |                | DP3g95maf05.fil1.vcf
| step 6         | filter out sites with reads from both strands     | SKIP                 | **SKIP**      |                | SKIP
| step 7         | ration mapping qualities ref vs alternate alleles | 297                  | 9689          |                | DP3g95maf05.fil3.vcf
| step 8         | paired status                                     | 297                  | 9689          |                | DP3g95maf05.fil4.vcf
| step 9         | remove sites quality score < 1/4 depth            | 297                  | 9688          |                | DP3g95maf05.fil5.vcf
| step 10        | depth x quality score cutoff	                     | 297                  | 8325          | 11.00          | 
| step 11        | He > 0.6 & Fis > 0.5 & Fix < -0.5                 | 297                  | 8206          | 27 min         | DP3g95maf05.FIL.HFis.recode.vcf
| step 12        | rename                                            |                      |               |                | dip_all_filtered.vcf 


## 02-Bayescan

Detect Fst outliers by bayesian inference with the [BayeScan software](http://cmpg.unibe.ch/software/BayeScan/)

### step 1: convert vcf files to Bayescan .txt files

This script will load your vcf file, determine the population identifier for each 
individual, and return a bayescan .txt inputfile
eboulanger's avatar
eboulanger committed
73
74
75
76
77
set arguments :    
  $1 = input file (vcf)  
  $2 = species code  
The script is currently set to detect and assign two populations (K). Run the script 
interactively if you want to determine and assign other values of K.
eboulanger's avatar
eboulanger committed
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

#### for mullus :
```
Rscript --vanilla Bayescan_input.R ../01-SNPfilters/02-Mullus/mul_all_filtered.vcf mul
```

#### for diplodus :
```
Rscript --vanilla Bayescan_input.R ../01-SNPfilters/01-Diplodus/dips_all_filtered.vcf dip
```

### step 2: determine outliers with Bayescan

download and compile Bayescan from [here](http://cmpg.unibe.ch/software/BayeScan/download.html) 
and copy the executable to your local bin folder:
```
cp ~/programmes/BayeScan2.1/binaries/BayeScan2.1_macos64bits /usr/local/bin/
```
#### run the BayeScan model for mullus and diplodus data
multiple runs with different parameters and different input data, with two chains per run 
to compare convergence later
```
bash run_bayescan.sh
```
### step 3: verify convergence and extract outliers
Run interactive R script called `Bayescan_evaluation.R`
   run 1 seems to get "best" results for diplodus. Neither runs detects outliers for mullus.

The script also extracts outlier lists and export loci positions for later subsetting

## 03-PCAdapt

detect outliers with [PCAdapt](https://bcm-uga.github.io/pcadapt/articles/pcadapt.html) in 
R and retrieve loci positions from vcf file

### Run the PCAdapt.R script

This script will load your vcf file, detect outlier loci using the package pcadapt and
output a list of outlier loci positions that can be used to subset your vcf file

eboulanger's avatar
eboulanger committed
118
119
120
set arguments :  
$1 = input file (vcf)  
$2 = species code  
eboulanger's avatar
eboulanger committed
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141

#### for mullus :
```
Rscript --vanilla PCAdapt.R ../01-SNPfilters/02-Mullus/mul_all_filtered.vcf mul
```
see how many outliers were detected :
```
wc -l outl_pos_pcadpt_mul.txt
```
2632 outliers detected for mullus

#### for diplodus :
```
Rscript --vanilla PCAdapt.R ../01-SNPfilters/01-Diplodus/dip_all_filtered.vcf dip

wc -l outl_pos_pcadpt_dip.txt
```
388 outliers detected for diplodus

## 04-finalPrep

eboulanger's avatar
eboulanger committed
142
Final steps to get neutral and adaptive SNP sets and correct file formats for both species.
eboulanger's avatar
eboulanger committed
143

eboulanger's avatar
eboulanger committed
144
145
146
147
This script subsets the filtered vcf file from 01-SNPfilters by outlier positions detected 
in 02-BayeScan and 03-PCAdapt.  
It also subsets the same vcf file for the remaining neutral positions and applies a final 
filter for HWE.
eboulanger's avatar
eboulanger committed
148

eboulanger's avatar
eboulanger committed
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
Finally, the script converts the final adaptive and neutral .vcf files in .tped, .tfam, 
.bed and .raw format necessary for downstream analyses.

set arguments:  
  $1 = input file (vcf)  
  $2 = species code  
  
#### for diplodus
```
bash outlier_positions.sh ../01-SNPfilters/01-Diplodus/dip_all_filtered.vcf dip
```
#### for Mullus
```
bash outlier_positions.sh ../01-SNPfilters/02-Mullus/mul_all_filtered.vcf mul
```
In total, 2680 adaptive loci were detected, with 10 loci detected by both the BayeScan and PCAdapt method.
After HWE filter, 12432 neutral loci were retained.

#### clean up files to separate directories
```
mkdir 01-Diplodus
mv dip_* 01-Diplodus/

mkdir 02-Mullus
mv mul_* 02-Mullus
```