seqbias_fit: Fitting seqbias models
In seqbias: Estimation of per-position bias in high-throughput sequencing data

Description Usage Arguments Details Value Note Author(s) See Also Examples

Fits a seqbias module given a reference sequence and reads in BAM format

1	seqbias.fit(ref_fn, reads_fn, n = 1e5, L = 15, R = 15)

`ref_fn`	filename of a reference sequence against which the reads are aligned, in FASTA format.
`reads_fn`	filename of aligned reads in BAM format.
`n`	train on at most this many reads.
`L`	consider at most L positions to the left of the read start.
`R`	consider at most R positions to the right of the read start.

A Bayesian network is trained on the first n unique reads in the provided BAM file, predicting the posterior probability of a read beginning at a position given the surrounding sequence. This is used to discern the sequencing bias: how more or less likely a read is to fall on a particular position.

The abundance of region can be more accurately assessed by normalizing (dividing) each position by its predicted bias.

A vector of reals giving the predicted sequencing bias for each position.

Both the BAM file and the FASTA file should be indexed, with, 'samtools index' and, 'samtools faidx' respectively.

Daniel Jones dcjones@cs.washington.edu

seqbias.predict

  reads_fn <- system.file( "extra/example.bam", package = "seqbias" )
  ref_fn <- system.file( "extra/example.fa", package = "seqbias" )

  sb <- seqbias.fit( ref_fn, reads_fn )

  I <- GRanges( c('seq1'), IRanges( c(1), c(5000) ), strand = c('-') )

  bias <- seqbias.predict( sb, I )