Skip to main content

Table 2 Different methods were used in the hackathons and are also available as reproducible vignettes

From: Community-wide hackathons to identify central themes in single-cell multi-omics

Common challenges Tasks Hackathon 1 (spatial transcriptomics) Hackathon 2 (spatial proteomics) Hackathon 3 (scNMT-seq)
Pre-processing Normalization & data transformation Data distribution checks (Fig. 2: 1, Fig. 2: 4)
High Variable Genes selection (Fig. 2: 5)
Variance Stabilization Normalization [16] (Fig. 2: 8)
Arcsinh transformation (Fig. 2: 9).
Inverse transformation (Fig. 2: 9)
Selection of patients (Fig. 2: 9)
Summaries of DNA measurements (input data provided in hackathon)
Managing differences in scale Data integration LIGER [17] (Fig. 2: 2) (sc)
ComBat (Fig. 2: 4) (bulk)
Projection methods MFA, sGCCA [18] (Fig. 2: 4a) (bulk)
UMAP/tSNE (Fig. 2: 2) (sc)
Multiblock PCA [19]
Weighting matrices based on their similarities: STATIS, MFA (Fig. 2: 8) (bulka)
Scale MIBI-TOF to the range of CyTOF values (Fig. 2: 9)
LIGER [17] (Fig. 2: 13) (sc)
Projection method sGCCA [18] (Fig. 2: 11) (bulk)
Multi Omics Supervised Integrative Clustering with weights (Fig. 2: 14) (bulk)
Overlap Cell overlap
(features not matching)
   Dimension reduction and projection methods:
LIGER [17] (Fig. 2: 13) (sc)
sGCCA [18] (Fig. 2: 11) (bulk)
Partial feature overlap
(cells not matching)
  Imputation:
Direct inversion with latent variables
Optimal transport to predict protein expression (Fig. 2: 10)
K-nearest neighbor averaging (Fig. 2: 9)
No imputation:
Biological Network Interactiona
 
Partial cell overlap
(features not matching)
  Multiblock PCA [19] (Fig. 2: 8a)  
No cell overlap
(complete feature overlap)
  Transfer cell type label with Random Forest (Fig. 2: 7) LIGER [17] (Fig. 2: 13)
No cell overlap
(partial feature overlap)
  Topic modeling to predict cell spatial co-location or spatial expression (Fig. 2: 9, partial feature overlap)  
No overlap   RLQa [20]  
Generic approaches Classification & feature selection Backward selection with SVM (Fig. 2: 1)
Self-training ENet (Fig. 2: 4)
Balanced error rate (Fig. 2: 1) Fig. 2: 4)
Recursive Feature Elimination (Fig. 2: 5)
(all bulk)
  Multi Omics Supervised Integrative Clustering (Fig. 2: 14) (bulk)
Lasso penalization in regression-type models (bulk)
Cell type prediction Projection with LIGER [17] (Fig. 2: 2)
SVM (Fig. 2: 1, Fig. 2: 5)
ssEnet (Fig. 2: 4)
(all bulk)
  
Spatial analysis Hidden Markov random field
Voronoi tesselation (Fig. 2: 1) (bulk)
Spatial autocorrelation with Moran’s Index (Fig. 2: 7, Fig. 2: 10)
Selection of spatial discriminative features:
Moran’s Index, NN correlation, Cell type, interaction composition, L function (Fig. 2: 10)
(all bulk)
 
Inclusion of additional information   Survival prediction: Cox regression based on spatial features (Fig. 2: 10) Include annotated hypersensitive sites index to anchor new/unseen data from DNase-seq, (sc)ATAC-seq, scNMT-seq, for de novo peak calling (bulka)
  1. aindicates that the method was not applied on the hackathon data, “bulk” indicates the method was originally developed for bulk omics, “sc” indicates the method was specifically developed for single-cell data, other methods are generic