Fig. 5: scLong outperformed existing scRNA-seq foundation models and task-specific methods in gene regulatory network inference. | Nature Communications

Fig. 5: scLong outperformed existing scRNA-seq foundation models and task-specific methods in gene regulatory network inference.

From: scLong: a billion-parameter foundation model for capturing long-range gene context in single-cell transcriptomics

Fig. 5: scLong outperformed existing scRNA-seq foundation models and task-specific methods in gene regulatory network inference.

a Model architecture for fine-tuning the pretrained scLong for this inference task. b scLong achieves higher area under the precision-recall curve ratio (AUPR) and early precision ratio (EPR) compared to Geneformer, scGPT, scFoundation, UCE, and task-specific DeepSEM and GENIE3. In (b), bar heights represent the mean and error bars indicate the standard deviation across n = 5 independent training runs with different random seeds. Results from individual runs are shown as dot points. Source data are provided as a Source Data file. Two-sided t-tests with Benjamini-Hochberg correction were used; see Supplementary Table 14 for detailed statistics.

Back to article page