Testing scClassify model — predict_scClassify (original) (raw)
Testing scClassify model
predict_scClassify( exprsMat_test, trainRes, cellTypes_test = NULL, k = 10, prob_threshold = 0.7, cor_threshold_static = 0.5, cor_threshold_high = 0.7, features = "limma", algorithm = "WKNN", similarity = "pearson", cutoff_method = c("dynamic", "static"), weighted_ensemble = FALSE, weights = NULL, parallel = FALSE, BPPARAM = BiocParallel::SerialParam(), verbose = FALSE )
Arguments
| exprsMat_test | A list or a matrix indicates the log-transformed expression matrices of the query datasets |
|---|---|
| trainRes | A `scClassifyTrainModel` or a `list` indicates scClassify trained model |
| cellTypes_test | A list or a vector indicates cell types of the qurey datasets (Optional). |
| k | An integer indicates the number of neighbour |
| prob_threshold | A numeric indicates the probability threshold for KNN/WKNN/DWKNN. |
| cor_threshold_static | A numeric indicates the static correlation threshold. |
| cor_threshold_high | A numeric indicates the highest correlation threshold |
| features | A vector indicates the gene selection method, set as "limma" by default. This should be one or more of "limma", "DV", "DD", "chisq", "BI". |
| algorithm | A vector indicates the KNN method that are used, set as "WKNN" by default. This should be one or more of "WKNN", "KNN", "DWKNN". |
| similarity | A vector indicates the similarity measure that are used, set as "pearson" by default. This should be one or more of "pearson", "spearman", "cosine", "jaccard", "kendall", "binomial", "weighted_rank","manhattan" |
| cutoff_method | A vector indicates the method to cutoff the correlation distribution. Set as "dynamic" by default. |
| weighted_ensemble | A logical input indicates in ensemble learning, whether the results is combined by a weighted score for each base classifier. |
| weights | A vector indicates the weights for ensemble |
| parallel | A logical input indicates whether running in paralllel or not |
| BPPARAM | A BiocParallelParam class object from the BiocParallel package is used. Default is SerialParam(). |
| verbose | A logical input indicates whether the intermediate steps will be printed |
Value
list of results
Examples
data("scClassify_example") wang_cellTypes <- scClassify_example$wang_cellTypes exprsMat_wang_subset <- scClassify_example$exprsMat_wang_subset data("trainClassExample_xin")
pred_res <- predict_scClassify(exprsMat_test = exprsMat_wang_subset, trainRes = trainClassExample_xin, cellTypes_test = wang_cellTypes, algorithm = "WKNN", features = c("limma"), similarity = c("pearson"), prob_threshold = 0.7, verbose = TRUE)
#> Ensemble learning is disabled... #> Using parameters: #> similarity algorithm features #> "pearson" "WKNN" "limma" #> [1] "Using dynamic correlation cutoff..." #> [1] "Using dynamic correlation cutoff..." #> classify_res #> correct correctly unassigned intermediate #> 0.704590818 0.239520958 0.000000000 #> incorrectly unassigned error assigned misclassified #> 0.000000000 0.051896208 0.003992016 #> weights for each base method: #> numeric(0)