SPaRTAN (Single-cell Proteomic and RNA based Transcription factor Activity Network) provides a mechanistically inspired approach for integrating cell-specific transcriptomic and proteomic data with regulatory genomics resources, representing a significant advance in the modeling of cell-specific signaling and gene regulatory programs. The cell surface phenotype is well-known to immunologists through flow cytometry but signaling downstream of cell surface receptors/co-receptors drives transcriptional and chromatin state changes. It is important to connect the “cell surface phenotype” to downstream transcriptional programs and resulting transcriptomic phenotypes. SPaRTAN models this flow of information with single cell resolution. The description of the method and some usage examples are available in https://doi.org/10.1101/2020.12.22.423961. SPaRTAN greatly enhances the utility of CITE-seq datasets to uncover TF and cell-surface receptor relationships in diverse cellular states.
Briefly, our model views the expression of surface proteins as a proxy of their activities; signaling emanating from these proteins converges on particular TFs, whose activities, in turn, regulate the expression of their target genes. Specifically, we use a regularized bilinear regression algorithm called affinity regression (AR), a general statistical framework for any problem where the observed data can be explained as interactions (W) between two kinds of inputs, to establish an interaction matrix between surface proteins (P) and TFs (D) that predicts target gene expression (Y).
Since the model captures statistical relationships between surface proteins, TFs, and gene expression. We can use the trained interaction matrix to obtain different views of a CITE-seq data set; e.g., to predict TF activity from a cell's surface protein expression profile or to predict surface protein expression from a cell’s gene expression profile. Intuitively, information flows down from observed surface protein levels through the learned interaction matrix to infer TF activities and observed mRNA expression levels or propagates up through the TF target-gene edges and interaction network to infer surface protein expression.
Currently we have implementations of SPaRTAN in MATLAB as well as in Python. The output from any of the implementations can then be explored either in R or Python. Since Python is an open souce package, no license needed, here we give detailed instructions of installation and usage of python implementation. Please see Run SPaRTAN in Python
Ma X*, Somasundaram A*, Qi Z, Singh H, Osmanbeyoglu HU, SPaRTAN, a computational framework for linking cell-surface receptors to transcriptional regulators. bioRxiv 2020.12.22.423961: doi:https://doi.org/10.1101/2020.12.22.423961
Pelossof, R., Singh, I., Yang, J.L., Weirauch, M.T., Hughes, T.R. and Leslie, C.S. (2015) Affinity regression predicts the recognition code of nucleic acid-binding proteins. Nat Biotechnol, 33, 1242-1249.
Osmanbeyoglu, H.U., Pelossof, R., Bromberg, J.F. and Leslie, C.S. (2014) Linking signaling pathways to transcriptional programs in breast cancer. Genome Res, 24, 1869-1880.
Osmanbeyoglu, H.U., Toska, E., Chan, C., Baselga, J. and Leslie, C.S. (2017) Pancancer modelling predicts the context-specific impact of somatic mutations on transcriptional programs. Nat Commun, 8, 14249.
Garcia-Alonso, L., Holland, C.H., Ibrahim, M.M., Turei, D. and Saez-Rodriguez, J. (2019) Benchmark and integration of resources for the estimation of human transcription factor activities. Genome Res, 29, 1363-1375.