Introduction
Platform-Agnostic CellNet (PACNet) is the newest version of CellNet. Just like CellNet, PACNet is a network-biology-based, computational platform that assesses the fidelity of cellular engineering and generates hypotheses for improving cell derivations. However, PACNet is applicable to many transcriptome profiling methods, expression estimations methods, and it does not require 100% gene overlap between query and reference samples. These features enable cross-study comparisons of cell fate engineering protocols. We have harvested panels of expression data of engineered cells from state-of-the-field protocols for HSPC, heart, intestine/colon, liver, lung, neuron, and skeletal_muscle. This means that you can compare your engineered cells to the derived from the most commonly used protocols without having to re-do those experiments yourself. The PACNet web app is only available for human data. However, we provide the resources and instructions for applying PACNet human and mouse here. Below, we describe how to use this web app to analyze human bulk-derived RNA-Seq data.
How to use this application
Inputs for CellNet Analysis:
1) Study name : desired study name
2) .csv sample metadata table with the following columns at minimum: sample_name, description1
Please make sure the columns are named exactly as above. Extra columns will be ignored.
3) .csv counts, TPM, or FPKM expression matrix with gene symbols as row names and sample names as column names. COLUMN NAMES OF EXPRESSION MATRIX MUST MATCH sample_name COLUMN OF SAMPLE METADATA TABLE. Column order does not matter.
4) Target cell or tissue type. If your target cell/tissue type is not one of those listed, please select "none_of_the_above" as your Target Cell/Tissue Type. In this case, neither comparison against engineered reference samples nor calculation of Network Influence Scores will be performed; however, PACNet will still produce classification results for your data.
Browsing reference data:
Alternatively, choose 'Browse reference data' to browse available engineered reference datasets rather analyzing an input dataset.
Outputs:
1) Classification heatmap : Columns represent query samples, and rows represent tissue types of the training data. Each square is colored by the query sample's classification score for the given tissue type. Scores range from 0 (distinct from the tissue type of the training data) to 1 (indistinguishable from the tissue type of the training data).
Classification heatmaps for panels of engineered reference samples are available for HSPC, heart, intestine/colon, liver, lung, neuron, and skeletal_muscle and will be displayed automatically alongside your results.
2) Network Influence Score (NIS) for transcriptional regulators : The transcriptional regulators of the tissue GRN are shown on the y axis, with the NIS of each regulator on the x axis. The NIS prioritizes transcription factors (TFs) such that their experimental perturbation is predicted to improve the target tissue classification. Negative scores suggest that upregulation of a given TF's expression is needed to achieve the target cell type, and positive scores suggest that downregulation is needed.
If there are more than 50 TFs in a cell/tissue type-specific network, then NIS scores for only the top 50 TFs by absolute value of NIS score will be displayed. However, the full matrix of NIS scores for each TF in the cell/tissue type-specific network and each sample is available for download.