fastTopics wrapper

.Rbuildignore

@@ -1,3 +1,5 @@
 ^.*\.Rproj$
 ^\.Rproj\.user$
 ^docs$
+^NOTES\.txt$
+

DESCRIPTION

@@ -23,6 +23,7 @@ Remotes: welch-lab/liger,
     cole-trapnell-lab/monocle3,
     mojaveazure/seurat-disk,
     powellgenomicslab/Nebulosa,
+    stephenslab/fastTopics,
     atakanekiz/CIPR-Package,
     prabhakarlab/Banksy
 Depends:
@@ -45,11 +46,12 @@ Collate:
     'internal.R'
     'alevin.R'
     'alra.R'
-    'banksy.R' 
+    'banksy.R'
     'cellbrowser.R'
     'cogaps.R'
     'conos.R'
     'fast_mnn.R'
+    'fasttopics.R'
     'glmpca.R'
     'liger.R'
     'miqc.R'
@@ -59,7 +61,7 @@ Collate:
     'velocity.R'
 Encoding: UTF-8
 LazyData: true
-RoxygenNote: 7.1.1
+RoxygenNote: 7.2.3
 Suggests:
     cipr,
     conos,
@@ -74,6 +76,7 @@ Suggests:
     fishpond,
     monocle3,
     CoGAPS,
+    fastTopics,
     glmpca,
     Nebulosa,
     presto,

NAMESPACE

@@ -8,6 +8,9 @@ S3method(as.Seurat,list)
 S3method(as.cell_data_set,Seurat)
 export(ALRAChooseKPlot)
 export(ExportToCellbrowser)
+export(FitPoissonNMF)
+export(FitTopicModel)
+export(PerformGoMDEAnalysis)
 export(PlotMiQC)
 export(ReadAlevin)
 export(ReadVelocity)
@@ -29,6 +32,8 @@ export(StopCellbrowser)
 export(as.cell_data_set)
 importFrom(BiocManager,install)
 importFrom(Matrix,Matrix)
+importFrom(Matrix,colSums)
+importFrom(Matrix,t)
 importFrom(Matrix,writeMM)
 importFrom(R.utils,gzip)
 importFrom(Seurat,"DefaultAssay<-")
@@ -52,6 +57,7 @@ importFrom(Seurat,Idents)
 importFrom(Seurat,IsGlobal)
 importFrom(Seurat,Loadings)
 importFrom(Seurat,LogSeuratCommand)
+importFrom(Seurat,Misc)
 importFrom(Seurat,Project)
 importFrom(Seurat,Reductions)
 importFrom(Seurat,SelectIntegrationFeatures)
@@ -66,6 +72,9 @@ importFrom(cowplot,theme_cowplot)
 importFrom(data.table,data.table)
 importFrom(data.table,fwrite)
 importFrom(data.table,setDTthreads)
+importFrom(fastTopics,de_analysis)
+importFrom(fastTopics,fit_poisson_nmf)
+importFrom(fastTopics,fit_topic_model)
 importFrom(ggplot2,aes_string)
 importFrom(ggplot2,geom_line)
 importFrom(ggplot2,geom_point)
@@ -86,6 +95,7 @@ importFrom(rlang,duplicate)
 importFrom(rsvd,rsvd)
 importFrom(stats,as.dist)
 importFrom(stats,pnorm)
+importFrom(stats,prcomp)
 importFrom(stats,quantile)
 importFrom(stats,sd)
 importFrom(stats,setNames)

NOTES.txt

@@ -0,0 +1,25 @@
+Comments from Alan Selewa:
+
+I was able to run the wrapper on my heart single cell RNA-seq
+data. Theres about ~70k cells and 26k genes. Fitting the model took
+about 4 hours using 16 Rcpp parallel threads. Heres some thoughts on
+the process:
+
+FitTopicModel would benefit from more hyper-parameters in its
+interface, like number of EM iterations to perform, number of parallel
+threads to use, etc. Basically anything you’d pass to fastTopics
+should be available in the interface, in my opinion.
+
+I would add a function that works on the Seurat object to plot
+log-likelihood vs iteration (similar to Seurat::ElbowPlot).
+
+It seems that downstream analyses (structure_plot, differential
+expression) will work directly on the fit produced. In conformity with
+everything else Seurat-related, I think these functions should work
+directly on the Seurat object, rather than extracting the fit object,
+and running functions found in fastTopics.
+
+The topic proportion of each cell should be visualized with
+Seurat::FeaturePlots e.g. topic 1 proportion overlayed on a UMAP as I
+shown in the vignette below add a wrapper function for visualizing
+gene proportion in each topic.