> ## Documentation Index
> Fetch the complete documentation index at: https://mintlify.com/satijalab/seurat-wrappers/llms.txt
> Use this file to discover all available pages before exploring further.

# Conos Integration

> Joint graph construction for mapping between multiple single-cell RNA-seq datasets.

Conos (Clustering On Network Of Samples) builds a joint graph across multiple single-cell datasets by finding and weighting mutual nearest neighbors in a shared embedding space. SeuratWrappers provides an `as.Seurat()` conversion method for Conos objects, allowing you to bring integrated results back into the Seurat ecosystem for downstream analysis.

## Citation

If you use Conos in your work, please cite:

> *Joint analysis of heterogeneous single-cell RNA-seq dataset collections*
>
> Nikolas Barkas, Viktor Petukhov, Daria Nikolaeva, Yaroslav Lozinsky, Samuel Demharter, Konstantin Khodosevich, Peter V. Kharchenko
>
> Nature Methods, 2019
>
> doi: [10.1038/s41592-019-0466-z](https://doi.org/10.1038/s41592-019-0466-z)
>
> GitHub: [https://github.com/hms-dbmi/conos](https://github.com/hms-dbmi/conos)

## Installation

```r theme={null}
# Install Conos from GitHub
remotes::install_github('hms-dbmi/conos')

# Install SeuratWrappers for the as.Seurat conversion method
remotes::install_github('satijalab/seurat-wrappers')
```

## Workflow

<Steps>
  <Step title="Load libraries and data">
    ```r theme={null}
    library(conos)
    library(Seurat)
    library(SeuratData)
    library(SeuratWrappers)

    InstallData("pbmcsca")
    data("pbmcsca")
    ```
  </Step>

  <Step title="Preprocess each dataset independently">
    Split the merged Seurat object and run standard preprocessing on each subset separately. Conos works on a list of preprocessed Seurat objects.

    ```r theme={null}
    pbmcsca.panel <- SplitObject(pbmcsca, split.by = "Method")
    for (i in 1:length(pbmcsca.panel)) {
      pbmcsca.panel[[i]] <- NormalizeData(pbmcsca.panel[[i]]) %>%
        FindVariableFeatures() %>%
        ScaleData() %>%
        RunPCA(verbose = FALSE)
    }
    ```
  </Step>

  <Step title="Build a Conos object">
    Pass the list of preprocessed Seurat objects to `Conos$new()`.

    ```r theme={null}
    pbmcsca.con <- Conos$new(pbmcsca.panel)
    ```
  </Step>

  <Step title="Build the joint graph">
    Connect datasets by finding mutual nearest neighbors in PCA space.

    ```r theme={null}
    pbmcsca.con$buildGraph(
      k = 15,
      k.self = 5,
      space = "PCA",
      ncomps = 30,
      n.odgenes = 2000,
      matching.method = "mNN",
      metric = "angular",
      score.component.variance = TRUE,
      verbose = TRUE
    )
    ```
  </Step>

  <Step title="Find communities and embed">
    Run community detection and compute a joint graph embedding.

    ```r theme={null}
    pbmcsca.con$findCommunities()
    pbmcsca.con$embedGraph()
    ```
  </Step>

  <Step title="Convert to Seurat and visualize">
    Use the SeuratWrappers `as.Seurat()` method to convert the Conos object back to a Seurat object. The joint embedding is stored as the `largeVis` reduction.

    ```r theme={null}
    pbmcsca <- as.Seurat(pbmcsca.con)
    DimPlot(pbmcsca, reduction = "largeVis",
            group.by = c("Method", "ident", "CellType"), ncol = 3)
    ```
  </Step>
</Steps>

## Examples

### Interferon-stimulated and control PBMC

```r theme={null}
InstallData("ifnb")
data("ifnb")
ifnb.panel <- SplitObject(ifnb, split.by = "stim")
for (i in 1:length(ifnb.panel)) {
  ifnb.panel[[i]] <- NormalizeData(ifnb.panel[[i]]) %>%
    FindVariableFeatures() %>%
    ScaleData() %>%
    RunPCA(verbose = FALSE)
}
ifnb.con <- Conos$new(ifnb.panel)
ifnb.con$buildGraph(
  k = 15, k.self = 5, space = "PCA", ncomps = 30,
  n.odgenes = 2000, matching.method = "mNN",
  metric = "angular", score.component.variance = TRUE, verbose = TRUE
)
ifnb.con$findCommunities()
ifnb.con$embedGraph()
ifnb <- as.Seurat(ifnb.con)
DimPlot(ifnb, reduction = "largeVis",
        group.by = c("stim", "ident", "seurat_annotations"), ncol = 3)
```

### Eight human pancreatic islet datasets

```r theme={null}
InstallData("panc8")
data("panc8")
panc8.panel <- SplitObject(panc8, split.by = "replicate")
for (i in 1:length(panc8.panel)) {
  panc8.panel[[i]] <- NormalizeData(panc8.panel[[i]]) %>%
    FindVariableFeatures() %>%
    ScaleData() %>%
    RunPCA(verbose = FALSE)
}
panc8.con <- Conos$new(panc8.panel)
panc8.con$buildGraph(
  k = 15, k.self = 5, space = "PCA", ncomps = 30,
  n.odgenes = 2000, matching.method = "mNN",
  metric = "angular", score.component.variance = TRUE, verbose = TRUE
)
panc8.con$findCommunities()
panc8.con$embedGraph()
panc8 <- as.Seurat(panc8.con)
DimPlot(panc8, reduction = "largeVis",
        group.by = c("replicate", "ident", "celltype"), ncol = 3)
```

## SeuratWrappers Function: as.Seurat.Conos

SeuratWrappers registers an `as.Seurat()` S3 method for Conos objects. When called on a Conos object after `embedGraph()`, it:

* Combines all cells from all input datasets into a single Seurat object
* Transfers cell metadata (sample labels, community assignments) into the Seurat object
* Stores the Conos joint embedding (e.g., `largeVis` or `UMAP`) as a `DimReduc` object
* Propagates community cluster identities to `Idents()`

```r theme={null}
# After building and embedding the Conos graph:
seurat.obj <- as.Seurat(conos.obj)
```

<Note>
  The reduction name in the converted Seurat object depends on which embedding method was used in `embedGraph()`. The default is `largeVis`. If you used UMAP embedding, it will appear as `UMAP`.
</Note>