Concerto

This repository contains code and tutorials for the following tasks: query-to-reference task, multimodal self-supervised pretraining task, batch correction task and Covid19 task.

Getting started

In the first step, you will need to download datasets to run each notebook and reproduce the result. The download links for datasets are shown in a folder named "data" in the following tasks directories.

Basic Usage (details see notebooks in each task)

from concerto_function5_3 import *

# preprocessing
adata = preprocessing_rna(adata, batch_key='tech')

# make tfrecord
ref_tf_path = concerto_make_tfrecord(adata_ref,tf_path = save_path + 'tfrecord/ref_tf/',batch_col_name = 'tech')
query_tf_path = concerto_make_tfrecord(adata_query,tf_path = save_path + 'tfrecord/query_tf/',batch_col_name = 'tech')

# integrate
concerto_train_ref_query(ref_tf_path,query_tf_path,weight_path)

# query to reference mapping
ref_embedding,query_embedding,ref_id,query_id = concerto_test_ref_query(weight_path,ref_tf_path,query_tf_path)

# NNvoting according to cells' embedding
query_neighbor,query_prob = knn_classifier(ref_embedding,query_embedding,adata_ref,ref_id,column_name='celltype',k=5)

Query-to-reference task (Fig4)

• Notebook path: transfer/tutorial_transfer.ipynb
• Readme for the transfer tutorial
• Description: Human pancreas dataset is used to perform query-to-reference mapping (HP->inDrop).

Multimodal self-supervised pretraining task (Fig3)

• Cell embeddings are generated by self-supervised pretraining and used for clustering: Multimodal_pretraining/tutorial_multimodal_cluster.ipynb
• Plot attention weight of self-supervised pretraining: Multimodal_pretraining/tutorial_multimodal_print_attention.ipynb
• Readme for the multimodal self-supervised pretraining tutorial
• Description: We use PBMC160K multimodal dataset to self-supervised pretrain a multimodal model, which can be used to perform clustering and extract attention weight.

Batch correction task (Fig3, extend Fig 1)

• Notebook path: Batch_correction/tutorial_overcorrect.ipynb
• Readme for the batch corrtection tutorial
• Description: To justify Concerto’s ability to avoid over-correction, we design a controlled experiment using a simulated dataset.

Covid19 task (Fig5)

• Note book path: Covid19_task/Covid19_demo.ipynb
• Readme for the Covid19 tutorial
• Description: Mapping Covid-19 cells against the integrated reference from 10X and DNBelab-C4 data.

Cross tissue annotation (Fig2g）

• Note book path: Cross_tissue_annotation/tutorial_inter.ipynb
• Readme for the cross-tissue annotation tutorial
• Description: Cross-tissue prediction on the TMS dataset.

Dataset used in this study with corresponding links（Supplementary Table 1&7）

• The download link of dataset is shown in SupplementaryTable1&7_links.xlsx