Artificial Neural Network System for Cell Classification using Single Cell RNA Expression

Abstract

We implemented an automated system for single-cell classification using artificial neural networks (ANN). Our system takes single-cell gene expression sparse matrices and trains ANN to classify cell types and subtypes. The assemblies of ANNs predict cell classes by voting. We tested the system in a case study where we trained ANNs with a dataset containing approximately 120,000 single cells and tested the resulting model using an independent data set of 13,000 single cells. The overall accuracy of the 5-class classification was 95%. We trained and tested a total of 100 ANNs in 10 cycles. The prediction system demonstrated excellent reproducibility. The analysis of misclassifications indicated that 2% were likely classification errors, while the remaining 3% were likely due to mislabeled types and subtypes in the test set.

Prediction of therapy outcomes of CLL using gene expression intensity, clustering, and ANN classification of single cell transcriptomes

“…For this analysis we used SCT classification of healthy PBMC by ANN that we developed earlier [29] to assess the SCT changes associated with successful CLL treatment.…”

“…The current study is the extension of our studies where we trained ANN to classify PBMC in five main classes. Our trained network reported 95% accuracy in classification of healthy PBMC into 5 main classes (positive predictive values were BC -92.0%, DC -94.3%, MC -97.0%, NK -81.5%, and TC -97.4%) [29]. We applied 5-class classification to CLL PBMC samples to observe the behavior of PBMC in CLL before and during therapy.…”

“…Standardization of data sets was performed as described earlier [9]. The ANN model was developed as described in [9,29].…”

“…This trained ANN was used to assign the 5 main PBMC subtype to the CLL samples before and during the therapy. The ANN was assessed as 95% accurate in classification of five PBMC cell types (BC, DC, MC, NK, and TC) [29]. We used radar plots to compare different patients on the same therapy day, and also compared changes of cell type proportions for each patient, over time.…”

“…We defined a diverse reference data set for classification of healthy PBMC and trained an artificial neural network (ANN) to achieve 95% accuracy in 5-class classification of PBMC [29]. The five main cell classes in PBMC include B cells (BC), dendritic cells (DC), monocytes (MC), natural killer cells (NK), and T cells (TC).…”