scSemiProfiler.inference.tgtinfer

scSemiProfiler.inference.tgtinfer(name, representative, target, bulktype='pseudobulk', lambdad=4.0, pretrain1batch=128, pretrain1lr=0.001, pretrain1vae=100, pretrain1gan=100, lambdabulkr=1, pretrain2lr=0.0001, pretrain2vae=50, pretrain2gan=50, inferepochs=150, lambdabulkt=8.0, inferlr=0.0002, pseudocount=0.1, k=15, device='cuda:0')[source]

Computationally infer the single-cell data of a single non-representative target sample based on a representatives’ single-cell data and bulk data of both samples.

Parameters

  • name (str) – The project name.

  • representative (typing.Union[str, int]) – The representative. Either indicated using sample ID (str) or the i-th (int) sample.

  • target (typing.Union[str, int]) – The target sample. Either indicated using sample ID (str) or the i-th (int) sample.

  • bulktype (str) – Pseudobulk or real bulk data

  • lambdad (float) – Scaling factor for the discriminator loss.

  • pretrain1batch (int) – The mini-batch size during the first pretrain stage.

  • pretrain1lr (float) – The learning rate used in the first pretrain stage.

  • pretrain1vae (int) – The number of epochs for training the VAE during the first pretrain stage.

  • pretrain1gan (int) – The number of iterations for training GAN during the first pretrain stage.

  • lambdabulkr (float) – Scaling factor for represenatative bulk loss for pretrain 2.

  • pretrain2lr (float) – Pretrain 2 learning rate.

  • pretrain2vae (int) – The number of epochs for training the VAE during the second pretrain stage.

  • pretrain2gan (int) – The number of iterations for training the GAN during the second pretrain stage.

  • inferepochs (int) – The number of epochs used for each mini-stage during inference.

  • lambdabulkt (float) – Scaling factor for the initial target bulk loss.

  • inferlr (float) – Infer stage learning rate.

  • k (int) – The number of nearest neighbors used in cell graph.

  • device (str) – Which device to use, e.g. ‘cpu’, ‘cuda:0’.

  • pseudocount (float) – Pseudocount used when converting real bulk to pseudobulk space

Return type

None

Returns

None

Example

>>> name = 'project_name'
>>> scSemiProfiler.tgtinfer(name = name, representatives = 6, target = 7, bulktype = 'real')