nuget
A point set of high-dimensional observations, stripped of all single-cell biology semantics. Each sample is represented by a mean vector (coords) and a per-dimension standard deviation (uncertainty). This mirrors the Bonsai SCData abstraction but only keeps the pure numeric content required by the tree-reconstruction core.
PointSet
01 Syntax
Microsoft.VisualBasic.DataMining.Bonsai.PointSet
02 Methods
| Name | Overloads | Summary |
|---|---|---|
| EstimateGeneVariance | 1 | Per-dimension global variance v_g = (1/(N-1)) Σ_i (means(i,g) - x_g)^2, the empirical spread of each gene across all samples. |
| GetSNR | 1 | Signal-to-noise ratio of each dimension: S_g = (1/C) * Σ_i (means(i,g) - x_g)^2 / ε²_gi where ε²_gi is the measurement variance (stds squared) of cell i on gene g. |
| FilterBySNR | 1 | Return a copy of this point set that keeps only the dimensions whose signal-to-noise ratio S_g is at least threshold (Bonsai keeps S_g >= 1 by default). |
| FromMatrix | 1 | Build a point set directly from a flat row-major array. |
| GetMean | 1 | Extract the mean vector of a single sample as a D-length array. |
| GetVar | 1 | Extract the variance vector of a single sample as a D-length array. |
| GetStd | 1 | Convenience accessor for the standard deviation matrix. |
03 Properties
| Name | Overloads | Summary |
|---|---|---|
| useGlobalVariance | 1 | When true, the likelihood uses v_g * tParent (global gene variance scaled diffusion) instead of the plain measurement-error variance in the transition variance. |
04 Fields
| Name | Overloads | Summary |
|---|---|---|
| means | 2 | N x D matrix of sample means. |
| stds | 2 | N x D matrix of per-dimension standard deviations. |
| names | 2 | N sample labels (optional, used for Newick export and result annotation). |
| nSamples | 2 | Number of samples (rows). |
| nGenes | 2 | Number of dimensions per sample (columns). |
| vars | 2 | Per-dimension variances, i.e. |
| geneVariance | 2 | Per-dimension global gene variance v_g, estimated as the sample variance of each dimension across all cells: v_g = (1/(N-1)) * Σ_i (means(i,g) - x_g)^2. |
05 Members
EstimateGeneVariance(
Double()())Per-dimension global variance v_g = (1/(N-1)) Σ_i (means(i,g) - x_g)^2, the empirical spread of each gene across all samples. Mirrors the gene-variance prior used by Bonsai.
GetSNR(
Double)Signal-to-noise ratio of each dimension: S_g = (1/C) Σ_i (means(i,g) - x_g)^2 / ε²gi where ε²gi is the measurement variance (stds squared) of cell i on gene g. Following the Bonsai paper, only dimensions with S_g >= threshold* are kept for tree construction.
FilterBySNR(
Double)Return a copy of this point set that keeps only the dimensions whose signal-to-noise ratio S_g is at least threshold (Bonsai keeps S_g >= 1 by default). Returns the original set unchanged when every dimension passes.
FromMatrix(
Double()(), Double()(), String())Build a point set directly from a flat row-major array.
GetMean(
Int32)Extract the mean vector of a single sample as a D-length array.
GetVar(
Int32)Extract the variance vector of a single sample as a D-length array.
GetStd(
Int32)Convenience accessor for the standard deviation matrix.
useGlobalVariance
When true, the likelihood uses v_g * tParent (global gene variance scaled diffusion) instead of the plain measurement-error variance in the transition variance. Default off, so existing numerical results are preserved unless explicitly enabled.
means
N x D matrix of sample means. Row i is the D-dimensional coordinate of sample i.
stds
N x D matrix of per-dimension standard deviations. PointSet.stds(i, g) is the uncertainty on PointSet.means(i, g). When nothing is known these default to 1.
names
N sample labels (optional, used for Newick export and result annotation).
nSamples
Number of samples (rows).
nGenes
Number of dimensions per sample (columns).
vars
Per-dimension variances, i.e. PointSet.stds squared. Pre-computed for speed.
geneVariance
Per-dimension global gene variance v_g, estimated as the sample variance of each dimension across all cells: v_g = (1/(N-1)) * Σ_i (means(i,g) - x_g)^2. Used as the optional diffusion prior (which assumes the diffusion magnitude along dimension g is proportional to v_g) described in the Bonsai paper (equation 4). When PointSet.useGlobalVariance is off, the likelihood reverts to the default behaviour of using the per-cell measurement error only.
means
stds
names
nSamples
nGenes
vars
geneVariance