Component: iSeq

iSeq

Wraps the iSeq R package that implements the methods described in "A fully Bayesian hidden Ising model for ChIP-seq data analysis" by Qianxing Mo

Version	1.0
Bundle	sequencing
Categories
Authors	Lauri Lyly (lauri.lyly@helsinki.fi)
Issue tracker	View/Report issues
Requires	iSeq (R-bioconductor)
Source files	component.xml iSeq.r
Usage	Example with default values iSeq( treatment = CSV, control = CSV, a0 = 1, a1 = 0.5, b0 = 1, b1 = 1, burnin = 500, ctcut = 0.95, cutoff = 0.5, gap = 300, k0 = 3, maxgap = 300, maxk = 10, maxlen = 80, method = "iSeq1", mink = 0, minlen = 10, normsd = 0.1, ntagcut = 10, peakreg_method = "ppcut", sampling = 2000, verbose = false )

Version

1.0

Bundle

sequencing

Categories

Authors

Lauri Lyly (lauri.lyly@helsinki.fi)

Issue tracker

View/Report issues

Requires

iSeq (R-bioconductor)

Source files

component.xml iSeq.r

Usage

Example with default values

Inputs

Name	Type	Mandatory	Description
treatment	CSV	Mandatory	Sample CSV file, with fields: chromosome name, region middle position, strand (either 1 for forward or 2 for reverse)
control	CSV	Mandatory	Control CSV file with similar format as treatment file.

Name

Type

Mandatory

Description

treatment

CSV

Mandatory

Sample CSV file, with fields: chromosome name, region middle position, strand (either 1 for forward or 2 for reverse)

control

CSV

Mandatory

Control CSV file with similar format as treatment file.

Outputs

Name	Type	Description
peaks	CSV	Called peaks regions returned from peakreg: A data frame with rows corresponding to enriched regions and columns corresponding to the following: chr: Chromosome IDs. gstart: The start genomic position of the enriched region. gend: The end genomic position of the enriched region. rstart: The row number for gstart in chrpos. rend: The row number for gend in chrpos. peakpos: The inferred center (peak) of the enriched region. meanpp: The mean posterior probability of the merged regions/bins. ct1: total tag counts for the region from gstart to gend for the chain corresponding to count[,1]; ct1=sum(count[rstart:rend,1]) ct2: total tag counts for the region from gstart to gend for the chain corresponding to count[,1]; ct2=sum(count[rstart:rend,2]) ct12: ct12 = ct1 + ct2 sym: A parameter used to measure if the forward and reverse tag counts are symmetrical (or balanced) in enriched regions. The values range from 0.5 (perfect symmetry) to 0 (complete asymmetry).
report	Latex	Various plots. To be determined more exactly - no plot is yet produced.

Name

Type

Description

peaks

CSV

Called peaks regions returned from peakreg: A data frame with rows corresponding to enriched regions and columns corresponding to the following: chr: Chromosome IDs. gstart: The start genomic position of the enriched region. gend: The end genomic position of the enriched region. rstart: The row number for gstart in chrpos. rend: The row number for gend in chrpos. peakpos: The inferred center (peak) of the enriched region. meanpp: The mean posterior probability of the merged regions/bins. ct1: total tag counts for the region from gstart to gend for the chain corresponding to count[,1]; ct1=sum(count[rstart:rend,1]) ct2: total tag counts for the region from gstart to gend for the chain corresponding to count[,1]; ct2=sum(count[rstart:rend,2]) ct12: ct12 = ct1 + ct2 sym: A parameter used to measure if the forward and reverse tag counts are symmetrical (or balanced) in enriched regions. The values range from 0.5 (perfect symmetry) to 0 (complete asymmetry).

report

Latex

Various plots. To be determined more exactly - no plot is yet produced.

Parameters

Name	Type	Default	Description
a0	float	1	a0: The scale hyper-parameter of the Gamma prior, alpha0.
a1	float	0.5	a1: The scale hyper-parameter of the Gamma prior, alpha1.
b0	float	1	b0: The rate hyper-parameter of the Gamma prior, beta0.
b1	float	1	b1: The rate hyper-parameter of the Gamma prior, beta1.
burnin	int	500	burnin: The number of MCMC burn-in iterations.
ctcut	float	0.95	ctcut: A value used to set the initial state for each window/bin. If tag count of a bin is greater than quantile(Y[,4],probs=ctcut), its state will be set to 1, otherwise -1. For typical ChIP-seq data, because the major regions are non-enriched, a good value for ctcut could be in the interval (0.9, 0.99).
cutoff	float	0.5	The cutoff value (a scalar) used to call enriched bins. If use posterior probability as a criterion (method="ppcut"), a bin is said to be enriched if its pp is greater than the cutoff. If use FDR as a criterion (method="fdrcut"), bins are said to be enriched if the bin-based FDR is less than the cutoff. The FDR is calculated using a direct posterior probability approach (Newton et al., 2004). The default value 0.5 is applicable to ppcut, a more likely default for fdrcut would be 0.05.
gap	int	300	gap: gap is the average length of the sequenced DNA fragments. If the distance between two nearest bins is greater than 'gap', a bin with 0 tag count is inserted into the two bins for modeling.
k0	float	3	k0: The initial parameter used to control the strength of interaction between neighboring bins, which must be a positive value (k0>0). A larger value of kappa represents a stronger interaction between neighboring bins.
maxgap	int	300	The criterion used to merge enriched bins. If the genomic distance of adjacent bins is less than maxgap, the bins will be merged into the same enriched region.
maxk	float	10	Unused for iSeq2. maxk: The maximum value of k(kappa) allowed.
maxlen	int	80	The maximum length of the genomic window/bin into which sequence tags are aggregated.
method	string	"iSeq1"	Either iSeq1 or iSeq2. iSeq1 implements the method that models the bin-based tag counts using Poisson-Gamma distribution and the hidden states of the bins using a standard 1D Ising model. iSeq2 is similar but uses a hidden high-order Ising model.
mink	float	0	Unused for iSeq2. mink: The minimum value of k(kappa) allowed.
minlen	int	10	The minimum length of the genomic window/bin into which sequence tags are aggregated.
normsd	float	0.1	Unused for iSeq2. normsd: iSeq1 uses a Metropolis random walk proposal for sampling from the posterior distributions of the model parameter kappa. The proposal distribution is a normal distribution with mean 0 and standard deviation specified by normsd
ntagcut	int	10	The tag count cutoff value for triggering bin size change. For example, suppose L_i and C_i are the length and tag count for bin i, respectively. If C_i >= ntagcut, the length for bin i+1 will be min(L_i/2,minlen); if C_i < ntagcut, the length for bin i+1 will be max(2L_i, maxlen). Note, by default, the bin sizes decrease/increase by a factor of 2. Thus, the user should let maxlen = (2^n)minlen.
peakreg_method	string	"ppcut"	'ppcut' or 'fdrcut', depending on whether cutoff is applied to posterior probability values or false discovery rate.
sampling	int	2000	sampling: The number of MCMC sampling iterations. The posterior probability of enriched and non-enriched state is calculated based on the samples generated in the sampling period.
verbose	boolean	false	verbose: A logical variable. If TRUE, the number of completed MCMC iterations is reported.

Name

Type

Default

Description

float

a0: The scale hyper-parameter of the Gamma prior, alpha0.

float

0.5

a1: The scale hyper-parameter of the Gamma prior, alpha1.

float

b0: The rate hyper-parameter of the Gamma prior, beta0.

float

b1: The rate hyper-parameter of the Gamma prior, beta1.

burnin

int

500

burnin: The number of MCMC burn-in iterations.

ctcut

float

0.95

ctcut: A value used to set the initial state for each window/bin. If tag count of a bin is greater than quantile(Y[,4],probs=ctcut), its state will be set to 1, otherwise -1. For typical ChIP-seq data, because the major regions are non-enriched, a good value for ctcut could be in the interval (0.9, 0.99).

cutoff

float

0.5

The cutoff value (a scalar) used to call enriched bins. If use posterior probability as a criterion (method="ppcut"), a bin is said to be enriched if its pp is greater than the cutoff. If use FDR as a criterion (method="fdrcut"), bins are said to be enriched if the bin-based FDR is less than the cutoff. The FDR is calculated using a direct posterior probability approach (Newton et al., 2004). The default value 0.5 is applicable to ppcut, a more likely default for fdrcut would be 0.05.

gap

int

300

gap: gap is the average length of the sequenced DNA fragments. If the distance between two nearest bins is greater than 'gap', a bin with 0 tag count is inserted into the two bins for modeling.

float

k0: The initial parameter used to control the strength of interaction between neighboring bins, which must be a positive value (k0>0). A larger value of kappa represents a stronger interaction between neighboring bins.

maxgap

int

300

The criterion used to merge enriched bins. If the genomic distance of adjacent bins is less than maxgap, the bins will be merged into the same enriched region.

maxk

float

Unused for iSeq2. maxk: The maximum value of k(kappa) allowed.

maxlen

int

The maximum length of the genomic window/bin into which sequence tags are aggregated.

method

string

"iSeq1"

Either iSeq1 or iSeq2. iSeq1 implements the method that models the bin-based tag counts using Poisson-Gamma distribution and the hidden states of the bins using a standard 1D Ising model. iSeq2 is similar but uses a hidden high-order Ising model.

mink

float

Unused for iSeq2. mink: The minimum value of k(kappa) allowed.

minlen

int

The minimum length of the genomic window/bin into which sequence tags are aggregated.

normsd

float

0.1

Unused for iSeq2. normsd: iSeq1 uses a Metropolis random walk proposal for sampling from the posterior distributions of the model parameter kappa. The proposal distribution is a normal distribution with mean 0 and standard deviation specified by normsd

ntagcut

int

The tag count cutoff value for triggering bin size change. For example, suppose L_i and C_i are the length and tag count for bin i, respectively. If C_i >= ntagcut, the length for bin i+1 will be min(L_i/2,minlen); if C_i < ntagcut, the length for bin i+1 will be max(2*L_i, maxlen). Note, by default, the bin sizes decrease/increase by a factor of 2. Thus, the user should let maxlen = (2^n)*minlen.

peakreg_method

string

"ppcut"

'ppcut' or 'fdrcut', depending on whether cutoff is applied to posterior probability values or false discovery rate.

sampling

int

2000

sampling: The number of MCMC sampling iterations. The posterior probability of enriched and non-enriched state is calculated based on the samples generated in the sampling period.

verbose

boolean

false

verbose: A logical variable. If TRUE, the number of completed MCMC iterations is reported.

Test cases