Compare TALEs by predicted DNA-binding specificity (FuncTAL)

Quantifies how TALE arrays relate by the DNA sequence their repeats are predicted to bind. (tales_compare_distal compares them by domain sequence instead.) The function turns each array's repeats into a position weight matrix (PWM) of the bases its RVDs prefer, and compares the PWMs pairwise with compare_motifs.

Usage

tales_compare_functal(
  x,
  method = "PCC",
  tryRC = FALSE,
  min.overlap = 1,
  normalise.scores = TRUE,
  score.strat = "a.mean",
  nthreads = 1
)

Arguments

x

A tales object carrying an rvd column.

method

One of compare_motifs()'s comparison metrics ("PCC" default). "PCC", "WPCC", "SW", "ALLR", "ALLR_LL" and "BHAT" are similarities (higher is more similar); the rest are already distances. Either kind is turned into a proper dissim in the result.

tryRC

Also score each pair's reverse-complement alignment and keep the better one. FALSE by default: an RVD array's specificity code has a fixed reading direction (5' to 3' target, N- to C-terminal repeat order), so comparing against a reverse complement asks a different, narrower biological question – do these two TALEs target opposite strands of overlapping sites. Ask it deliberately, when it is the question.

min.overlap

Minimum aligned width to accept, as compare_motifs() defines it. Defaults to 1 (any overlap at all); compare_motifs()'s own default is 6, suited to generic transcription-factor motifs, but a TALE array can have as few as half a dozen repeats, and that default would silently refuse to compare some real, short arrays.

normalise.scores

Penalise alignments that leave much of either motif unaligned. TRUE by default: TALE array lengths vary widely (see min.overlap), so an unnormalised score would favour matches that are mostly one array hanging off the end of a much longer one.

score.strat

How compare_motifs() combines per-column scores into one alignment score. "a.mean" (its own default): a sum would scale with array length, favouring long-vs-long comparisons for no biological reason.

nthreads

Passed to compare_motifs().

Value

A tale_distances object, interchangeable with tales_compare_distal's, so it can be handed directly to tales_group_hclust/tales_group_kmedoids. No tree is built here, matching tales_compare_distal()'s own division of labour: compare here, cluster/tree there.

Details

This is a reimplementation of QueTAL's FuncTAL comparison on universalmotif; the original Perl tool depends on BioPerl's Bio::Perl module, which current BioPerl no longer provides. Results differ from the original FuncTAL tool, by design. FuncTAL scored two RVD arrays by flattening their entire padded, overlapping alignment (positions and bases together) into one vector and taking a single Pearson correlation. compare_motifs() instead correlates matched columns individually and combines the column scores (score.strat); the two disagree on real data.

tales_to_universalmotif builds the PWMs (see its page for what drives them and how it treats an RVD missing from rvd_dna_specificity). It is exported so that its output can go to any universalmotif function.

Only a handful of compare_motifs's many options are exposed here, chosen for what actually varies across TALE arrays (their differing repeat counts); see ?compare_motifs for the rest.

See also

tales_compare_distal(), comparing by domain sequence instead; tales_to_universalmotif(), the conversion step this uses; called directly, it lets any other universalmotif function ( motif_tree(), view_motifs(), scan_sequences(), merge_motifs(), ...) can be run on the same PWMs.

Other pairwise distances: [.pairwise_distances(), as.matrix.pairwise_distances(), distances_assert_square(), distances_restrict(), is_pairwise_distances(), pairwise_distances(), rvd_dna_specificity, tales_assign_domain_codes(), tales_compare_distal(), tales_domain_distances(), tales_group_hclust(), tales_group_kmedoids(), tales_tale_distances(), validate_pairwise_distances()

Examples

x <- tales_from_telltales(system.file("extdata", "tellTaleExampleOutput",
                                      package = "tantale"))
fn <- tales_compare_functal(x)
fn
#> # A tibble: 16 × 3
#>    id1       id2       dissim
#>    <chr>     <chr>      <dbl>
#>  1 ROI_00001 ROI_00001  0    
#>  2 ROI_00002 ROI_00001  0.763
#>  3 ROI_00003 ROI_00001  0.785
#>  4 ROI_00004 ROI_00001  0.714
#>  5 ROI_00001 ROI_00002  0.763
#>  6 ROI_00002 ROI_00002  0    
#>  7 ROI_00003 ROI_00002  0.696
#>  8 ROI_00004 ROI_00002  0.728
#>  9 ROI_00001 ROI_00003  0.785
#> 10 ROI_00002 ROI_00003  0.696
#> 11 ROI_00003 ROI_00003  0    
#> 12 ROI_00004 ROI_00003  0.712
#> 13 ROI_00001 ROI_00004  0.714
#> 14 ROI_00002 ROI_00004  0.728
#> 15 ROI_00003 ROI_00004  0.712
#> 16 ROI_00004 ROI_00004  0