DECODER-SC: A single-cell algorithm for quantifying transcription factor-target regulatory discordance

Existing transcription factor (TF) activity inference methods quantify regulatory potential but do not evaluate whether inferred TF activity is concordant with downstream target-gene expression. To address this limitation, we present DECODER-SC, a single-cell algorithm that quantifies TF-target regulatory discordance from single-cell RNA-sequencing data. DECODER-SC integrates direction-aware regulatory logic using DoRothEA activation and repression annotations, weighted discordance scoring, and permutation-based statistical inference to compute TF-level, interaction-level, and cell-level discordance metrics. Benchmarking against AUCell, VIPER, and TF-target correlation baseline demonstrated that discordance profiles capture regulatory structure not explained by activity magnitude or simple expression co-variation. Ablation analysis supported the contribution of activator/repressor-aware modeling to discordance detection, while robustness analyses demonstrated highly stable TF rankings across multiple threshold settings. Orthogonal Perturb-seq validation demonstrated a reproducible IRF1 discordance signal across experimental units (effect = 0.0643, 95% CI = 0.0441-0.0845, BH-adjusted P = 0.0235). Leave-one-target-out and VIPER-based sensitivity analyses confirmed that discordance classifications are robust to target-level circularity and alternative TF-activity inference (97.76% and 73.42% agreement, respectively). Replicate-aware sample-level analysis further showed that cell-level proliferation and pathway associations were exploratory rather than statistically independent. Application of DECODER-SC across multiple single-cell transcriptomic datasets identified recurrent regulatory discordance patterns associated with malignant progression, stress-adaptive transcriptional programs, and transitional cellular states. DECODER-SC expands current single-cell regulatory analysis by enabling systematic quantification of TF-target discordance across heterogeneous cellular populations.