TL;DR
This paper presents a novel computational approach to identify core promoter elements in human and mouse genomes by leveraging sequence conservation between orthologous genes. By combining positional weight matrices with cross-species conservation, they significantly improve the accuracy of core promoter element detection and reveal synergistic interactions between these elements, offering a more robust method for predicting transcription start sites.
What Problem Does This Paper Address?
Identifying functional regulatory elements in DNA, particularly core promoter elements that are crucial for initiating gene transcription by RNA polymerase II, is a significant challenge. These elements (like TATA box, Initiator, DPE, BRE, and MTE) have highly degenerate sequences, leading to a high rate of false positives when using standard motif-finding techniques. While experimental data can confirm these elements, it's often limited. The paper aims to overcome the limitations of purely computational approaches by using comparative genomics to filter out non-functional elements, thereby increasing the reliability of core promoter identification.
Methods
The authors developed a method that combines two key strategies:
- Positional Weight Matrices (PWMs): They constructed PWMs for the five core promoter elements (BRE, TATA, Inr, MTE, DPE) based on known consensus sequences and, for TATA, by analyzing sequences from GenBank. These matrices are more sensitive than simple consensus searches.
- Comparative Genomics: They leveraged the fact that functional DNA elements are likely to be conserved between orthologous human and mouse genes. By comparing identified motifs in human promoters with their corresponding regions in mouse orthologs, they could filter out motifs that were likely false positives (i.e., not conserved).
To assess the robustness of their findings, they employed a bootstrapping technique, creating randomized promoter sequences with identical nucleotide frequencies but scrambled order, to serve as a negative control. They also used a smaller, highly curated dataset from the Eukaryotic Promoter Database (EPD) to validate their results.
Datasets
- Primary Dataset: 9,010 pairs of experimentally supported human and mouse orthologous promoter sequences. This dataset was derived from the OMGProm database.
- Validation Dataset: A subset of 624 sequences from the primary dataset that also matched entries in the Eukaryotic Promoter Database (EPD).
- Training/Matrix Construction: A set of 4,100 pairs of aligned human and mouse orthologous promoter sequences were used to build the PWMs.
- Negative Control: Randomized promoter sequences generated by shuffling nucleotides or dinucleotides from the real promoter sequences.
Key Insights and Ideas
- Conservation as a Filter: The core idea that sequence conservation between orthologous genes can dramatically improve the signal-to-noise ratio for identifying functional regulatory elements. Non-functional motifs are less likely to be conserved across species.
- Synergistic Element Combinations: The paper highlights that core promoter elements often work in concert. They identified specific combinations of elements (e.g., TATA-Inr-MTE and BRE-Inr-MTE) that appear synergistically, meaning their co-occurrence and specific spacing are more likely than random chance, especially when conservation is considered.
- Novel Synergistic Pairings: The identification of the BRE-Inr-MTE combination as a novel synergistic relationship, adding to previously suggested interactions.
- Refined TATA Box Frequency: The study suggests that TATA boxes are less prevalent in human promoters than historically believed, aligning with more recent findings.
Key Results
- Reduced False Positives: Incorporating the conservation criterion significantly reduced the number of identified core promoter elements, indicating a substantial decrease in false positives. For example, the number of identified TATA boxes dropped from 1848 to 1167 when conservation was applied.
- Quantified Element Frequencies: Based on their conserved findings, they estimated the prevalence of core promoter elements in human promoters: ~22% for BRE, ~17% for TATA, ~62% for Inr, ~57% for MTE, and ~12% for DPE.
- Significant Synergistic Pairs: They found that certain pairs of elements are significantly more likely to occur at a synergistic distance when conserved between mouse and human. For instance, TATA-MTE and Inr-MTE showed strong evidence of synergy.
- Novel Three-Way Synergy: The study identified two significant three-way synergistic combinations: BRE-Inr-MTE (p = 4.1 × 10⁻⁶) and TATA-Inr-MTE (p = 0.0042).
Limitations and Future Work
- PWM Construction: The PWMs were not all based on experimentally verified data, and the TATA box PWM was derived from a mix of organisms and hand-annotated sequences, which might introduce some limitations.
- Dataset Bias: The dataset primarily consists of highly conserved orthologous genes, which might limit the generalizability of findings to less conserved promoters.
- De Novo Prediction: The authors acknowledge that their method, while improving accuracy, still suffers from a high false positive rate when used alone for de novo core promoter detection.
- Future Directions: They suggest that their approach, when combined with ab initio promoter prediction methods (which rely on sequence characteristics like dinucleotide frequencies), could lead to improved overall performance by mutually eliminating false positives. They also anticipate the discovery of novel core promoter elements.
Summary updated