Multi-Lineage Prediction Analysis with Schrödinger

Introduction

Single-cell RNA-Seq can be used to identify definitive specified progenitors, committed to a single lineage, as well as multipotential progenitor cells which display coincident expression of genes associated with distinct lineage programs. As recently demonstrated the algorithm ICGS in AltAnalyze can be used to identify both committed progenitor and mixed-lineage cell states, as well as associated gene expression programs in conjunction with the algorithm MarkerFinder. To formally define these states, a new algorithm called Schrödinger has been developed to; (i) identify predominantly lineage restricted markers from ICGS and MarkerFinder, coincidently expressed in alternative lineages, (ii) determine which cells in the evaluated or secondary test dataset express a lineage program (aggregate of markers from i), and (iii) determine which ICGS cell states are statistically enriched in cells with the strongest evidence of multilineage expression. These analyses are performed using the multiLineagePredict module of AltAnalyze. Note, that these methods are still in development.

Schrödinger produces multiple outputs from its analysis: 1. Heatmap of cells lineage binary scores (present vs. absent) for all cells. 2. t-SNE representation of the matrix for 1. 3. Multilineage coincident marker genes associated with each lineage 4. Lineage programs present in each cell

All Results are saved to the query dataset folder (see DataPlots for heatmaps).

Running Schrödinger Through the command-line

An introduction to running AltAnalyze on the commandline can be found here. Prior to running AltAnalyze make sure you install the latest species database (e.g., python AltAnalyze.py --update Official --species Mm --platform RNASeq --version 72). Examples of running ICGS using different single-cell platforms and options can be found in the above introductory link. Once ICGS is run, you can simply run Schrödinger with the below command:

python stats_scripts/multiLineagePredict.py
  --expdir "/ExpressionInput/exp.scRNA-Seq_normalized.txt"
  --m "/ExpressionOutput/MarkerFinder/AllGenes_correlations-ReplicateBased.txt"
  --ICGS "/ICGS/Clustering-exp.control-Guide3 Fos-hierarchical_cosine_correlation.txt"