StaVia (Via 2.0) is our new single-cell trajectory inference method that explores single-cell atlas-scale data and temporal and spatial studies enabled by. In addition to the full functionality of earlier versions, StaVia now offers (check out our paper for details)

1. Integration of metadata (e.g time-series labels, spatial coordinates): Using sequential metadata (temporal labels from longitudinal studies, hierarchical information from phylogenetic trees, spatial distances relevant to spatial omics data) to guide the cartography. Integrating RNA-velocity where applicable.
2. Higher Order Random Walks: Leveraging higher order random walks with memory to highlight key end-to-end differentiation pathways along the atlas
3. Atlas View: Via 2.0 offers a unique visualization of the predicted trajectory by intuitively merging the cell-cell graph connectivity with the high-resolution of single-cell embeddings. Visit the Gallery to see examples.
4. Generalizable and data modality agnostic StaVia still offers all the functionality of Via 1.0 across single-cell data modalities (scRNA-seq, imaging and flow cyometry, scATAC-seq) for types of topologies (disconnected, cyclic, tree) to infer pseudotimes, automated terminal state prediction and automated plotting of temporal gene dynamics along lineages.

StaVia extends the lazy-teleporting walks to higher order random walks with memory to allow better lineage detection, pathway recovery and preservation of global features in terms of computation and visualization. The cartographic approach combining high edge and spatial resolution produces informative and esthetically pleasing visualizations caled the Atlas View.

If you find our work useful, please consider citing our StaVia and Via 1.0 paper .

Tutorials and videos available on readthedocs with step-by-step code for real and simulated datasets. Tutorials explain how to generate cartographic visualizations for TI, tune parameters, obtain various outputs and also understand the importance of memory. Datasets (anndata h5ad) links are provided below.

✳️ You can start with the The tutorial/Notebook for multifurcating data which shows a step-by-step use case. ✳️

scATAC-seq dataset of Human Hematopoiesis represented by VIA graphs (click image to open interactive graph)

Refer to the Jupiter Notebooks to plot these fine-grained vector fields of the sc-trajectories even when there is no RNA-velocity available.

Please visit our readthedocs for the latest tutorials and videos on usage and installation

Dataset are available in the Datasets folder (smaller files) with larger datasets here.

We recommend setting up a new conda environment and reccomend python version 3.10. Versions 3.8 and 3.9 should also work. You can use the examples below, the Jupyter notebooks and/or the test script to make sure your installation works as expected.

conda create --name ViaEnv python=3.10 
pip install pyVIA // tested on linux Ubuntu 16.04 and Windows 10

This usually tries to install hnswlib, produces an error and automatically corrects itself by first installing pybind11 followed by hnswlib. To get a smoother installation, consider installing in the following order after creating a new conda environment:

pip install pybind11
pip install hnswlib
pip install pyVIA

git clone https://github.com/ShobiStassen/VIA.git 
python3 setup.py install // cd into the directory of the cloned VIA folder containing setup.py and issue this command

The pie-chart cluster-graph plot does not render correctly for MACs for the time-being. All other outputs are as expected.

conda create --name ViaEnv python=3.10 
pip install pybind11
conda install -c conda-forge hnswlib
pip install pyVIA

If the pip install doesn't work, it usually suffices to first install all the requirements (using pip) and subsequently install VIA (also using pip). Note that on Windows if you do not have Visual C++ (required for hnswlib) you can install using this link .

pip install pybind11, hnswlib, igraph, leidenalg>=0.7.0, umap-learn, numpy>=1.17, scipy, pandas>=0.25, sklearn, termcolor, pygam, phate, matplotlib,scanpy
pip install pyVIA

All examples and tests have been run on Linux and MAC OS. We find there are somtimes small modifications required to run on a Windows OS (see below). Windows requires minor modifications in calling the code due to the way multiprocessing works in Windows compared to Linux:

#when running from an IDE you need to call the function in the following way to ensure the parallel processing works:
import os
import pyVIA.core as via
f= os.path.join(r'C:\Users\...\Documents'+'\\')
def main():
    via.main_Toy(ncomps=10, knn=30,dataset='Toy3', random_seed=2,foldername= f)    
if __name__ =='__main__':
    main()
    
#when running directly from terminal:
import os
import pyVIA.core as via
f= os.path.join(r'C:\Users\...\Documents'+'\\')
via.main_Toy(ncomps=10, knn=30,dataset='Toy3', random_seed=2,foldername= f)