The continuous developments of single-cell RNA-Seq (scRNA-Seq) have sparked an immense interest in understanding intercellular crosstalk. Multiple tools and resources that aid the investigation of cell-cell communication (CCC) were published recently. However, these methods and resources are usually in a fixed combination of a tool and its corresponding resource, but in principle any resource could be combined with any method.
if (!requireNamespace("BiocManager", quietly = TRUE)) install.packages("BiocManager") if (!requireNamespace("remotes", quietly = TRUE)) install.packages("remotes") remotes::install_github('saezlab/liana')
See a tutorial how to use LIANA to run any combination of 7 methods, plus their consensus, and 16 different resources! The tutorial with the test data takes minutes to complete!
To this end, we built a framework to decouple the methods from their corresponding resources.
LIANA also goes a step further as it provides:
A robust and extendable architecture that aims to accelerate method development and benchmarks
A rank aggregate from the results of different methods
A customizable plethora of resources
The tools implemented in this repository are:
CellPhoneDBv2 (*, $)
NATMI (*, $)
edge_weights) (*, $)
LRscores) (SCA) (*, $)
iTALK-inspired 1-vs-rest LogFC score (
logfc_comb) (*, $)
cross-talk scores (*)
consensus_rank of the predictions is also provided using the RobustRankAggregate package
*The scoring systems from these methods were re-implemented in LIANA in order to account for multimeric complexes, to simplify the calls to the individual pipelines, or reduce any possible inconsistencies and redundancies in their downstream integration. If you wish to run LIANA with the original tools please see LIANA++.
$ Default methods in LIANA.
The following CCC resources are accessible via this pipeline:
LIANA’s default resource was generated from the
Consensus of several expert-curated resources, then filtered to additional quality control steps including literature support, complex re-union/consensus, and localisation.
All the resources above are retrieved from OmniPath, and more specifically OmnipathR. However, individual resources retrieved from the OmniPath web service are not to be affected by this, as each resource expected to be identical to its original form, apart from minor processing steps.
OmniPath itself serves as a composite CCC resource combining all the ones listed above + more.
OmniPath also collects further information about the roles and localisation of proteins in intercellular communication.
We made use of this information to generate the
Consensus resource. To obtain more information how we filtered the default
Consensus resource, as well as to explore custom filter options see customizing OmniPath resources.
If you are interested in making use of the LIANA architecture for your own method, this vignette provides instructions how to obtain a comprehensive table of LR statistics, which can then be used by custom scoring functions. In the same vignette are also instructions how to install and run the original methods via a convenient R wrapper, e.g. for their benchmark.
We appreciate any feedback, so please do not hesitate to open an issue on the liana github page!
Dimitrov, D., Türei, D., Garrido-Rodriguez M., Burmedi P.L., Nagai, J.S., Boys, C., Flores, R.O.R., Kim, H., Szalai, B., Costa, I.G., Valdeolivas, A., Dugourd, A. and Saez-Rodriguez, J. Comparison of methods and resources for cell-cell communication inference from single-cell RNA-Seq data. Nat Commun 13, 3224 (2022). https://doi.org/10.1038/s41467-022-30755-0
Türei, D., Valdeolivas, A., Gul, L., Palacio‐Escat, N., Klein, M., Ivanova, O., Ölbei, M., Gábor, A., Theis, F., Módos, D. and Korcsmáros, T., 2021. Integrated intra‐and intercellular signaling knowledge for multicellular omics analysis. Molecular systems biology, 17(3), p.e9923. https://doi.org/10.15252/msb.20209923