Simulate B or T cell receptor sequences by variational autoencodes(VAEs) trained with experimental data.

Source: R/Echidna_vae_generate.R

Simulate B or T cell receptor sequences by variational autoencodes(VAEs) trained with experimental data.

Echidna_vae_generate(
  sequence,
  n.train,
  n.sample,
  batch.size,
  latent.dim,
  intermediate.dim,
  epochs,
  epsilon.std,
  null.threshold
)

Arguments

sequence

a vector of seuqnece the model to be trained on

n.train

number of sequence to be used in training set, the rest will be in testing set

n.sample

number of new sequence to generate from VAE model

batch.size

set to larger to save time, set to smaller to same computing power

latent.dim

parameter used in VAE model

intermediate.dim

parameter used in VAE model

epochs

parameter used in VAE model

epsilon.std

parameter used in VAE model

null.threshold

threshold of predicted value to be considered as an existing base, default is 0.05. When generated sequence is too short, lower this threshold.

Value

A simulated VDJ repertoire on the basis of the input experimental repertoire