imap 1.0.0

The integration of single-cell RNA-sequencing datasets from multiple sources is critical for deciphering cell-cell heterogeneities and interactions in complex biological systems. We present a novel unsupervised batch removal framework, called iMAP, based on two state-of-art deep generative models – autoencoders and generative adversarial networks.

iMAP - Integration of multiple single-cell datasets by adversarial paired transfer networks

Installation

1. Prerequisites

• Install Python >= 3.6. Typically, you should use the Linux system and install a newest version of Anaconda or Miniconda .
• Install pytorch >= 1.1.0. To obtain the optimal performance of deep learning-based models, you should have a Nivdia GPU and install the appropriate version of CUDA. (We tested with CUDA = 9.0)
• Install scanpy >= 1.5.1 for pre-processing.
• (Optional) Install SHAP for interpretation.

2. Installation

The iMAP python package is available for pip install(pip install imap). The functions required for the stage I and II of iMAP could be imported from “imap.stage1” and “imap.stage2”, respectively.

Tutorials

Tutorials and API reference are available in the tutorials directory.