Input Format¶
Barcode File¶
A barcode file describes the individuals that can be contained in a data set, as well as their barcodes and spacer sequences. The file contains three blocks of information as tab separated values:
A mapping of p5 barcode index to p5 barcode sequence.
A mapping of p7 barcode index to p7 barcode sequence.
A mapping of p5 index and p7 index to p5 spacer, p7 spacer, individual name and further information.
Example:
# p5 barcodes
1 ATCACG
2 CGATGT
3 TTAGGC
4 TGACCA
# p7 barcodes
1 ATCACG
8 ACTTGA
# specimen
1 1 Individual 1 Annotation 1
3 1 AC Individual 2 Annotation 1
2 8 GAC AT Individual 3 Annotation 2
4 8 C AT Individual 4 Annotation 1
This file shows four individuals.
The individual Individual 2 is encoded using the the p5 barcode TTAGGC (index 3) and the p7 barcode ATCACG (index 1).
The p5 reads if this indiviual have a spacer of AC, while the p7 reads have an empty spacer sequence.
Any additional information, here only the generic Annotation 1, can be added separated by additional tabs.
It is possible to simulate different barcode lengths in the same dataset:
Example:
# p5 barcodes
1 ATCA
2 CGATGTAC
3 TTAGGCG
4 TGACCATGTACCG
# p7 barcodes
1 ATCACG
8 ACTTGATG
# specimen
1 1 Individual 1 Annotation 1
3 1 AC Individual 2 Annotation 1
2 8 GAC AT Individual 3 Annotation 2
4 8 C AT Individual 4 Annotation 1
The combination of barcode length and spacer length determines, how much of the simulated genomic sequence is discoverable in a read. Given the fixed read length, individuals with short barcodes have more free space for genomic sequence in the read setup than individuals with long barcodes. A set of example barcode files can be found in the bitbucket repository.
Note, that only individuals who share a p7 barcode can occur in one data set.
Quality Model¶
The quality model used by ddRAGE to sample quality values is saved as a probability vector for each read position. Consider a small example with three positions:
P(q=0) P(q=1) P(q=2)
pos 0 [0.0, 0.2, 0.8]
pos 1 [0.1, 0.3, 0.6]
pos 2 [0.5, 0.3, 0.2]
pos 3 [0.6, 0.3, 0.1]
The probabilities are saved as a 2d numpy array of doubles and read by ddRAGE using the function numpy.loadtxt.
You can specify your own quality model by creating a qmodel file. One way to achieve this using python is to create a numpy array and write it to a qmodel file using numpy.savetxt.
import numpy as np
nr_positions = 100
nr_quality_values = 104-33 # only consider valid PHRED scores in Illumina 1.8 format
values = np.zeros(shape=(nr_quality_values, nr_positions), dtype=np.double)
# fill in all probability values
for pos in range(nr_positions):
for phred_value in range(nr_quality_values):
# compute probability for the phred_value occuring at read position pos
values[phred_value][pos] = prob(phred_value, pos)
# save the file
np.savetxt("mymodel.qmodel", values)
If the desired read length (defined by -r parameter) surpasses the number of positions defined in the qmodel
the last value will be used to pad the probability matrix.
In this example with
P(q=0) P(q=1) P(q=2)
pos 0 0.0, 0.2, 0.8
pos 1 0.1, 0.3, 0.6
pos 2 0.5, 0.3, 0.2
pos 3 0.6, 0.3, 0.1
pos 4 0.6, 0.3, 0.1 padded from pos 3
... ...
pos n 0.6, 0.3, 0.1 padded from pos 3
All vectors after an empty or non-1 probability vector will be ignored and overwritten using the last valid entry:
P(q=0) P(q=1) P(q=2)
pos 0 0.0, 0.2, 0.8
pos 1 0.1, 0.3, 0.6
pos 2 0.0, 0.0, 0.0 this line does not sum up to 1
pos 3 0.6, 0.3, 0.1 this line will be ignored
pos 2 0.1, 0.3, 0.6 padded from pos 1
pos 3 0.1, 0.3, 0.6 padded from pos 1
pos 4 0.1, 0.3, 0.6 padded from pos 1
... ...
pos n 0.1, 0.3, 0.6 padded from pos 1