Skip to content

What is the ground truth for the dyngen data. #61

@cakeinspace

Description

@cakeinspace

My understanding is that dyngen generates count table by simulating a GRN. Usually when we measure counts from a scRNA-seq experiment. That has a poisson sample of the underlying counts of genes present in the cell. Now some people say that the biological and technical noise is Poisson and then you can define the ground truth as the underlying rate parameter of the poisson or you can say that the counts are sampled from a negative binomial in which case you have the rate and overdispersion parameter.

In dyngen, what is the definition of the ground truth values of the cell, because the counts that we get from dyngen have biological + technical noise added on to it.

I believe the ground truth is the reduced dimensional coordinate of the MDS embeddings of the pearson correlation matrix of the sampled counts with landmarks along the trajectory but given that the MDS embeddings will distort the true nearest neighbour distances between cells, it should be not treated as the ground truth. Hence It is unclear to me how to define the ground truth.

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions