predictTxFeaturesPerSample: Identification of splice junctions and exons from BAM file

Description Usage Arguments Details Value Author(s)

View source: R/features-prediction.R


Splice junctions and exons are predicted from genomic RNA-seq read alignments in BAM format.


predictTxFeaturesPerSample(file_bam, which, paired_end, read_length,
  frag_length, lib_size, min_junction_count, alpha, psi, beta, gamma,
  min_anchor, include_counts, retain_coverage, junctions_only, max_complexity,
  sample_name, verbose, cores)



BAM file with genomic RNA-seq read alignments


GRanges of genomic regions to be considered for feature prediction, passed to ScanBamParam


Logical, TRUE for paired-end data, FALSE for single-end data


Read length required for use with alpha


Fragment length for paired-end data required for use with alpha


Number of aligned fragments required for use with alpha


Minimum fragment count required for a splice junction to be included. If specified, argument alpha is ignored.


Minimum FPKM required for a splice junction to be included. Internally, FPKMs are converted to counts, requiring arguments read_length, frag_length and lib_size. alpha is ignored if argument min_junction_count is specified.


Minimum splice frequency required for a splice junction to be included


Minimum relative coverage required for an internal exon to be included


Minimum relative coverage required for a terminal exon to be included


Integer specifiying minimum anchor length


Logical indicating whether counts of compatible fragments should be included in metadata column “N”


Logical indicating whether coverage for each exon should be retained as an RleList in metadata column “coverage”. This allows filtering of features using more stringent criteria after the initial prediction.


Logical indicating whether predictions should be limited to identification of splice junctions only


Maximum allowed complexity. If a locus exceeds this threshold, it is skipped, resulting in a warning. Complexity is defined as the maximum number of unique predicted splice junctions overlapping a given position. High complexity regions are often due to spurious read alignments and can slow down processing. To disable this filter, set to NA.


Sample name used in messages


If TRUE, generate messages indicating progress


Number of cores available for parallel processing


For spliced alignments, the direction of transcription is inferred from the XS tag in the BAM file and used to assign strand information to the read, or fragment for paired-end data.

Feature prediction is performed in two steps. First, splice junctions are identified from spliced alignments. Second, exons are identified based on regions that are flanked by splice junctions and show sufficient coverage with compatible reads.

Splice junctions implied by read alignments are filtered based on fragment count and splice frequency. The splice frequency at the splice donor (acceptor) is defined as x_J/x_D (x_J/x_A), where x_J is the number of fragments containing the splice junction, and x_D (x_A) is the number of fragments overlapping the exon/intron (intron/exon) boundary. Fragments overlapping the spliced boundary can be either spliced or extend into the intron. To be included in predicted features, splice junctions must have fragment count at least min_junction_count or FPKM at least alpha, and splice frequency at both donor and acceptor at least psi.

Regions between any pair of identified splice junctions with sufficient compatible read coverage are considered candidate internal exons. Read coverage for a candidate exon is computed based on compatible fragments, i.e. fragments with matching (or missing) strand information and introns consistent with the exon under consideration. Candidate exons are included in predicted features if the minimum coverage is at least beta * number of junction-containing fragments for either flanking junctions.

Terminal exons are regions downstream or upstream of splice junctions with compatible fragment coverage at least gamma * number of junction-containing fragments.


TxFeatures object


Leonard Goldstein

SGSeq documentation built on Nov. 8, 2020, 8:31 p.m.