Evolutionary signatures for protein-coding genes

By studying evolution, we found this:

• Evolutionary signatures:  Reading Frame Conservation, Codon Substitution Biases, Ka/Ks rate
• Scaling to multiple species:  increased separation between genes and non-coding regions
• Combining these signatures:  Classification.

Revisiting the D. melanogaster protein-coding gene catalog

Purpose here is not to study gene evolution (see Eisen paper), but rather to leverage the 12 genomes to improve gene annotation quality.

• People involved:
• MIT (Manolis, Mike Lin)
• FlyBase (Bill Gelbart, Peili Zhang, and FlyBase-Harvard curators)
• BDGP (Sue Celniker, Joe Carlson)
• Identification of novel protein-coding sequence
• Method
• CONGO similar in effect to Siepel's Exoniphy, but uses more flexible discriminative algorithms
• Predictions
• Experimental results/validations
• BDGP August iPCR runs - new cDNAs in GenBank
• BDGP heterochromatin cDNAs (unpublished data)
• FB4.2 vs 4.3
• FB human curation (~400 annotation changes)
• Intersection with Affy transfrags (so-so enrichment)
• Mass spec (ETH Zurich/Sandra Lovenich, Erich Bruggner, Konrad Basler, Ernst Hafen)
• To Do:
• who are those genes (Dscam etc.), how many exons prior to this (alternative splicing!)
• is length of exon mult. of 3, or does it otherwise fit with splicing requirements
• What is the expected fraction of novel exons to be in an intron (1/3 ?)
• Dubious genes
• Method: identify genes where we can't find any comparative evidence to believe they are real, by multiple metrics and in multiple alignment sets
• Properties of set
• length/RFC distribution
• lack of GO terms, lack of names
• lack of cDNA/EST evidence
• many single exon/single ORF
• For some of those that are transcribed, we predict conserved noncoding elements in the transcripts (RNA genes, microRNA genes)
• "Confirmed" hypothetical genes
• Method
• Identify genes without cDNA/EST evidence, but strong evolutionary evidence in syntenic alignments
• Limited definition of "confirmed" (can be sure protein-coding, but can't really confirm gene structure, or identify alt. splicing variation)
• Corrections and adjustments to existing annotations
• Translation start adjustments: evolutionary evidence suggests translation starts at a downstream ATG
• Transcript model corrections: by detecting frameshifts near an intron
• ORF corrections: wrong ORF currently annotated
• Summary: revised gene catalog
• Incorporation in FlyBase

Discovery of non-canonical genic phenomena

• People involved:
• MIT (Manolis Kellis, Mike Lin)
• Harvard (Bill Gelbart, Andy Schroeder, others from FlyBase-Harvard)
• UCSC (Jakob Pedersen on RNA involvement in recoding)
• Translational readthrough: observe protein-coding signatures continuing straight past stop codon
• Frameshifts: observe adjacent windows conserved in different frames (not near an intron)
• Polycistronics/uORFs: observe well-conserved disjoint ORFs in known transcript models
  
  

Conserved non-coding regions

• People involved:
• MIT (Manolis Kellis, Mik Lin, Huy, Alex Stark, Pouya, Leo)
• Harvard (Bill Gelbart)
• CSHL (Greg Hannon, Julius Brennecke)
• Whitehead (Dave Bartel, Graham Ruby)
• UCSC (Jakob Pedersen)
• “ultraconserved” elements
• 1851 elements > 60nts 100% conserved in at least 11 species
• Enriched in intron/exon boundaries and intergenic regions
• Intron CNEs enriched for transcription factor genes
• intron/exon CNEs enriched in nervous system proteins/channels.
• Overlap with known enhancer elements?
• Blasts/Blast-pipeline (other species) is there (2days)
• Blast to Dmel for other fam. Members: data there 2 days)
• RNA genes
• tRNAs
• snoRNAs, snRNAs, rRNAs
• New types of RNA genes
• Secondary structure properties of mRNAs
• microRNAs: conservation-based identification of Drosophila miRNAs
• prediction selects against exons, transposon and repeat sequence
• top prediction have miRNA-like features not used for prediction
• top novel hairpins are validated by library cloning (Bartel, Hannon) with 90% accurary
• 28 novel mirnas validated, 9 prev. predicted are confirmed by clonining, 6 are corrected
• new miRNAs-family members, new miRNA families
• targets for new miRNAs
• miRNAs in the introns of msi, kis, E2F, cdc2D
• mature mirna 5'ends can be predicted with high accuracy, exceptions highlight importance of star sequence (though this is not a general trend!)
• prediction accuracy scales with branch-length, prediction of clade-specific miRNAs with high accurary is currently impossible
• estimate of (conserved) Drosophila miRNAs < 150

Gene regulation

• People involved
• MIT (Manolis Kellis, Alex Stark)
• Promoter motifs
• Properties of known regulatory motifs
• Signatures for motif discovery
• Computational validation (against known motifs, tissue-specific expression, GO, positional bias, no strand bias)
• 3’ UTR motifs
• Role in miRNA regulation
• Role in identification of new miRNA genes 
• Other elements Pumillo (PUF) binding sites (incl nanos)
• Identification of gene targets
• Transcriptional regulation
• Targets of miRNA genes
• Motif combinations and grammars
• Towards motif-based gene regulatory networks

Discussion:  Assessing power to identify functional elements with 12 genomes

• Evolutionary signatures
• Protein-coding genes
• miRNA genes
• motifs
• CNEs