TL;DR
The GENCODE consortium provides a comprehensive, high-quality reference annotation for the human and mouse genomes, crucial for accurate biological and clinical research. This paper details their ongoing efforts to integrate manual and computational methods, expanding and refining gene annotations, including novel transcripts and non-coding RNAs, to meet the growing demands of genomic analysis.
What Problem Does This Paper Address?
Accurate and detailed gene annotation is fundamental for understanding genome biology and driving clinical genomics. However, the complexity of genomes, with their intricate regulatory elements, alternative splicing, and non-coding RNAs, presents a significant challenge. Existing annotations can be incomplete, inconsistent, or lack the necessary detail for high-throughput analysis. The GENCODE project aims to address this by creating and maintaining a definitive reference annotation that integrates diverse evidence sources and employs rigorous validation methods, ensuring a foundational resource for the scientific community.
Methods
GENCODE employs a multi-pronged approach combining manual and computational annotation:
- Manual Annotation: This is the cornerstone, involving expert annotators who meticulously examine genomic regions. They use a "clone-by-clone" approach for comprehensive coverage and "targeted annotation" to address specific questions or regions flagged by computational analysis or external users. This method allows for nuanced decisions, such as distinguishing between protein-coding and pseudogenic loci, based on a holistic view of available evidence.
- Computational Annotation: A suite of computational tools and pipelines are used to identify and classify various genomic features. This includes:
- RNA Capture Long Seq (CLS): A novel method to capture and sequence full-length long non-coding RNAs (lncRNAs), providing high-resolution transcript models that significantly inform manual annotation.
- Proteomics Data Integration: Mass spectrometry datasets are used to validate and refine protein-coding gene annotations, with strict criteria applied to minimize false positives.
- Annotation Validation (RT-PCR-seq): A high-throughput method to experimentally test the quality of annotations, particularly for low-confidence transcribed loci and splice junctions.
- Comparative Genomics (PhyloCSF, CAT): Tools like PhyloCSF help distinguish protein-coding regions by analyzing evolutionary conservation, while the Comparative Annotation Toolkit (CAT) projects GENCODE annotations to related species, aiding in the annotation of laboratory mouse strains and great apes.
- Pseudogene Identification (Pseudopipe): A pipeline for identifying and annotating pseudogenes, classifying them by origin (processed, duplicated, unitary).
- Small Non-coding RNA (sncRNA) Annotation: Imported from databases like miRBase for miRNAs, and computationally identified for others like snoRNAs and scaRNAs using homology-based approaches.
- APPRIS: This database helps identify principal splice isoforms based on cross-species conservation and protein structure, providing a standardized way to represent the most functionally relevant transcripts.
Datasets
The GENCODE consortium utilizes a vast array of publicly available data and analysis, including:
- Transcriptomic data: From various sources, including RNA-seq and CAGE/PET tags.
- Proteomic data: From mass spectrometry experiments.
- Comparative genomics data: Alignments of multiple genome sequences.
- Public databases: Such as UniProt, RefSeq, miRBase, Rfam, and UCSC Genome Browser.
- Experimental data: Generated through methods like CLS and RT-PCR-seq.
The paper does not introduce new benchmark datasets but rather focuses on the continuous improvement and integration of existing data resources.
Key Insights and Ideas
- The Power of Integration: The success of GENCODE hinges on the seamless integration of manual curation with sophisticated computational pipelines. Neither approach alone is sufficient; their synergy is key to achieving high-quality, comprehensive annotations.
- The Growing Importance of Non-coding RNAs: The paper highlights the significant effort dedicated to annotating lncRNAs and sncRNAs, underscoring their increasing recognition as critical players in biological processes.
- The Dynamic Nature of Genomes: GENCODE annotations are not static. The continuous updates, driven by new experimental data and computational advancements, reflect the ongoing discovery and refinement of our understanding of gene structure and function.
- The Value of "Partial" Transcripts: The "GENCODE comprehensive" set includes partial transcripts, acknowledging that even incomplete splice forms can provide valuable biological insights and represent unique functional entities.
- Cross-Species Annotation: The development of tools like CAT demonstrates the utility of projecting human and mouse annotations to other species, facilitating comparative genomics and the study of evolutionary relationships.
Key Results
- GENCODE provides reference gene sets for both human and mouse, regularly updated and accessible through major genome browsers (Ensembl, UCSC) and FTP sites.
- The consortium has significantly expanded the annotation of lncRNAs and pseudogenes, particularly in the mouse genome.
- The CLS method has enabled the annotation of novel lncRNA transcripts at thousands of loci in human and mouse.
- RT-PCR-seq validation confirmed a high percentage of splice junctions (78% in human, 53% in mouse), demonstrating the accuracy of the annotations.
- Manual reannotation efforts have led to the addition of numerous novel protein-coding genes and pseudogenes, especially in human.
Limitations and Future Work
- Incompleteness: Despite continuous efforts, the authors acknowledge that the annotations are "not yet complete," particularly given the abundance of new transcriptomic and proteomic data that continues to emerge.
- "Lift37" Annotation: While versions of selected human releases projected from GRCh38 to GRCh37 are available, this "lift37" annotation set is not considered official reference annotation due to projection difficulties and only receives minimal support.
- Future Work: The GENCODE consortium plans to continue improving the quality and coverage of annotations by integrating cutting-edge developments in genome biology and leveraging new datasets. They aim to refine the representation of existing loci and expand coverage through the addition of novel loci and alternatively spliced transcripts. Future efforts will also include incorporating samples from a more diverse panel of tissues for CLS experiments.
Summary updated