Pre-Existing Heterogeneity Predicts Rare Proteostasis-Stress Programs Across Diverse Perturbations

Area of Science:

• Single-cell genomics
• Systems biology
• Computational biology

Background:

• Stress-associated transcriptional programs are frequently observed in single-cell perturbation studies.
• Their origin, whether stochastic or due to pre-existing cellular heterogeneity, is not well understood.
• These stress signals are often dismissed as technical noise.

Purpose of the Study:

• To develop a predictive framework for stress programs across diverse single-cell perturbation datasets.
• To investigate whether baseline cellular heterogeneity and perturbation identity predict stress responses.
• To determine if stress variation is a predictable biological signal rather than nuisance noise.

Main Methods:

• Constructed a cross-dataset prediction framework using 146,321 single cells and 926 perturbation tasks.
• Employed leave-task-out, leave-cell-line-out, and leave-perturbation-out cross-validation strategies.
• Utilized independent validation datasets, including bulk RNA-sequencing and donor-paired single-cell data, for external validation.

Main Results:

• Baseline cellular heterogeneity and perturbation identity accurately predicted integrated stress burden (R2=0.742, Pearson r=0.862).
• The prediction framework demonstrated generalization across tasks, cell lines, and perturbations, and retained signal across datasets.
• Independent validation confirmed stress axes, including unfolded protein response/integrated stress response (UPR/ISR) modules and heat shock/proteostasis patterns.
• Identified distinct stress program patterns: activator protein 1 (AP1) immediate, UPR/activating transcription factor 4 (ATF4), heat shock, replication-coupled, and low/mixed dominant.

Conclusions:

• Cellular stress variation is not merely nuisance noise but a predictable and biologically structured aspect of perturbation responses.
• Pre-existing cellular heterogeneity plays a significant role in determining stress outcomes post-perturbation.
• The developed framework and identified stress programs offer a new resource for analyzing perturbation responses in single-cell data.