Abstract
Our analysis identifies transcription factors associating with patient survival and provides insight into the regulatory mechanism underlying the breast cancer and leukemia. The transcription factors identified by our method are biologically meaningful and consistent with prior knowledge. As an insightful tool, this approach can also be applied to other microarray cancer data sets to help researchers better understand the intricate relationship between transcription factors and diseases.
Introduction
In cancer research, gene expression profiling via microarrays is a standard technique to identify dysregulated genes. However, the messenger RNA (mRNA) levels of transcription factors (TFs) often fail to reflect their actual regulatory activity. A TF may be highly expressed but inactive due to a lack of post-translational modifications or necessary co-factors. Conversely, a low-abundance TF might be highly efficient. This study addresses the need to understand TF activity—the actual influence on target genes—to better predict cancer progression and identify therapeutic targets.
Methodology
The researchers developed a computational framework utilizing the BASE algorithm to infer TF activity from downstream gene expression patterns. The process involved:
- Motif Integration: 565 Positional Weighted Matrices (PWMs) from the TRANSFAC database were used as TF binding "fingerprints."
- Binding Affinity Mapping: Instead of binary target lists, the binding affinity of each TF was calculated across the promoter regions (1kb upstream) of all human genes.
- Activity Inference: By integrating binding profiles with patient microarray data, the BASE algorithm generated an Activity Change (AC) score for each TF per sample, representing its functional impact.
- Survival Analysis: Spearman correlation and Cox proportional-hazards models identified TFs whose activity scores significantly correlated with patient survival in breast cancer and leukemia datasets.
Results and Discussion
The study demonstrated that inferred TF activity is a significantly more robust predictor of survival than raw mRNA expression. For example, while ATF4 mRNA levels showed weak correlation with outcomes, its inferred activity was highly significant.
Key Regulatory Drivers in Breast Cancer
In breast cancer, 26 PWMs were significantly associated with survival. Notably, 20 of these were "negative predictors," where higher TF activity correlated with shorter disease-free survival. The steroid nuclear receptor family (including Progesterone, Androgen, and Glucocorticoid receptors) emerged as central players.
Additionally, the ATF/CREB family was identified as a critical marker, particularly in ER-negative breast cancer subtypes. This suggests that different regulatory networks drive progression across various cancer subtypes.
Predictive Performance and Leukemia Findings
A logistic regression model using just four TF activities (PR, E2F, CREB, and PAX9) achieved 76% accuracy in predicting clinical outcomes. In leukemia, the activity of TAL1 was identified as a significant survival predictor in Acute Myeloid Leukemia (AML) patients. These findings remained consistent across different microarray platforms, validating the method's robustness.
Conclusion
While the current model focuses on proximal promoter regions and is limited by the noise inherent in PWMs, it provides a powerful tool for identifying "master regulators" of cancer. Future work integrating ChIP-seq data to map physical binding sites more accurately could further refine these activity scores. This framework offers a scalable approach to mapping the regulatory landscape of patient survival across diverse cancer types.
Summary updated