Find conserved and diverged target genes within a module — findConservedDivergedTargets • CroCoNet

Calculates a contribution score for each target gene in a conserved/diverged module and finds genes that are particularly responsible for the conservation/divergence.

Usage

findConservedDivergedTargets(
  module_name,
  tree_stats_jk,
  lm_tree_stats,
  conf_level = 0.95
)

Arguments

module_name

Character, the name of the module of interest.

tree_stats_jk

Data frame of tree statistics per jackknife module version across all modules.

Columns required in case the focus of interest is conservation and overall divergence:

regulator: Character, transcriptional regulator.
module_size: Integer, the numer of target genes assigned to a regulator.
type: Character, module type (orig = original or jk = jackknifed).
id: Character, the unique ID of the module version (format: nameOfRegulator_jk_nameOfGeneRemoved in case of module type 'jk' and nameOfRegulator_orig in case of module type 'orig').
gene_removed: Character, the name of the gene removed by jackknifing (NA in case of module type 'orig').
total_tree_length: Numeric, total tree length per jackknife module version.
within_species_diversity: Numeric, within-species diversity per jackknife module version.

Columns required in case the focus of interest is species-specific divergence:

regulator: Character, transcriptional regulator.
module_size: Integer, the numer of target genes assigned to a regulator.
type: Character, module type (orig = original or jk = jackknifed).
id: Character, the unique ID of the module version (format: nameOfRegulator_jk_nameOfGeneRemoved in case of module type 'jk' and nameOfRegulator_orig in case of module type 'orig').
gene_removed: Character, the name of the gene removed by jackknifing (NA in case of module type 'orig').
{{species}}_subtree_length: Numeric, the subtree length of a species per jackknife module version.
{{species}}_diversity: Numeric, diversity within the species of interest per jackknife module version.

lm_tree_stats

Object of class lm, the (weighted) linear model fit between the total tree length and within-species diversity (in case the focus of interest is conservation and overall divergence) or between the subtree length and diversity of a species (in case the focus of interest is the divergence between that particular species and all others).

conf_level

Confidence level of the prediction interval (default: 0.95).

Value

Data frame of cross-species conservation measures per target gene for the module of interest. In addition to the relevant columns of tree_stats_jk, it contains 6 new columns:

focus: Character, the focus of interest in terms of cross-species conservation, either "overall" if the focus of interest is conservation and overall divergence, or the name of a species if the focus of interest is the divergence between that particular species and all others.
fit: Numeric, the fitted total tree length/subtree length of a species at the diversity value of a jackknife module version.
lwr_fit: Numeric, the lower bound of the prediction interval of the fit.
upr_fit: Numeric, the upper bound of the prediction interval of the fit.
residual: Numeric, the residual of the jackknife module version in the linear model. It is calculated as the difference between the observed and expected (fitted) total tree length/subtree length of a species.
contribution: Numeric, the contribution score of the removed target gene.

It also contains a new summary row with the aggregate values across all jackknife values of the module.

Details

To determine whether a module as whole is conserved, diverged overall or diverged on a specific lineage, the CroConet approach relies on module trees reconstrcuted from pairwise preservation scores between replicates and statistics calculated based on these trees (total tree length, within-species diversity, subtree length and diversity of a particular species). To identify individual genes that contribute the most to conservation/divergence, the same statistics can be used in combination with jackknifing.

During jackknifing, each member gene of a module is removed and all statistics are recalculated, including the tree-based statistics that inform us about cross-species conservation (see the parameter jackknife in the function calculatePresStats). Our working hypothesis is that if removing a target from a diverged module makes the module more conserved, then that target was responsible for divergence in the original module, and vice versa, if removing a target from a conserved module makes the module more diverged, then that target was responsible for conservation in the original module.

To quantify these effects, the function takes use of the linear models that were fitted across all modules between the total tree length and within-species diversity (in case the focus of interest is conservation and overall divergence) or between the subtree length and diversity of a species (in case the focus of interest is species-specific divergence), similarly to how the conserved and diverged modules were identified in the first place. However, in this case it is not the aggregate statistic for a module as a whole that is compared to the regression line, but the statistic of each jackknife module version separately. Using the residual of a jackknifed version ($\hat{\epsilon}_i$) and of the original module ($\hat{\epsilon}_\text{orig}$), we defined the target gene contribution score ($c_i$) as follows: $$c_i = \frac{\hat{\epsilon}_\text{orig} - \hat{\epsilon}_i}{\hat{\epsilon}_\text{orig}}$$

This score is applicable both to conserved and to diverged modules: the highest score corresponds to the most diverged target gene in case of a diverged module and to the most conserved target gene in case of a conserved module. A score of zero means that removing a target gene does not affect the signal of conservation or divergence, i.e. the extent of cross-species differences associated with this gene reflect the module average. A negative score indicates that removing the target gene strengthens the signal of conservation or divergence, i.e. the target acts against the overall module trend. A positive score means that removing a target gene diminishes the signal of conservation or divergence, i.e. the target contributes to the observed signal when it is present. A score of 1 (which corresponds to $\hat{\epsilon}_i = 0$) means that the removal of the target gene completely eliminates the signal, therefore it is single-handedly responsible for the observed conservation or divergence. Finally, a score greater than 1 suggests that removing the gene reverses the direction of the signal, meaning that it is a strongly diverged gene within an otherwise conserved module, or vice versa.

It is important to note, that in most cases the divergence/conservation of a module cannot be contributed to a single target gene, but it is rather the combined signal of all targets together. Therefore contribution scores nearing or exceeding 1 are rare. This also means that the "most conserved" target gene of diverged module is still most likely diverged, just to a lesser extent than the others. Thus, it mainly makes sense to investigate the most diverged targets of the diverged modules and the most conserved targets of the conserved modules, not the other way around.

Based on the linear model provided, the function calculates the fitted values of the total tree length/species subtree length, the prediction interval of the fit, the residual of each jackknife module version and the original module, as well as the target gene contribution scores. The results can be plotted using the function plotConservedDivergedTargets.

Examples

POU5F1_target_conservation <- findConservedDivergedTargets("POU5F1", tree_stats_jk, lm_overall)