Discovery of high-confidence human protein-coding genes and exons by whole-genome PhyloCSF helps elucidate 118 GWAS loci

“Potential orthologs were initially sought using BLASTP (Altschul et al. 1990) on the vertebrate protein database atNCBI(NCBIResourceCoordinators2017)andexaminedintheir genomiccontextusingtheUCSC(Casperetal.2017)andEnsembl (Zerbino et al. 2018) Genome Browsers.”

“It has been almost two decades since the first high-quality sequences from the human genome became available (International Human Genome Sequencing Consortium 2001; Venter et al. 2001).”

“Multispecies protein alignments were created using Clustal Omega (Sievers et al. 2011).”

“It has been almost two decades since the first high-quality sequences from the human genome became available (International Human Genome Sequencing Consortium 2001; Venter et al. 2001).”

“At present, the two main human gene annotation projects, Ensembl/GENCODE (Harrow et al. 2012; Frankish et al. 2018; Zerbino et al. 2018) (henceforth GENCODE) and RefSeq (O’Leary et al. 2016), as well as the UniProt protein resource (The UniProt Consortium 2019), disagree on the number of human protein-coding genes (Ab…”

“At present, the two main human gene annotation projects, Ensembl/GEN-CODE (Harrow et al. 2012; Frankish et al. 2018; Zerbino et al. 2018) (henceforth GENCODE) and RefSeq (O ’ Leary et al. 2016), as well as the UniProt protein resource (The UniProt Consortium 2019), disagree on the number of human protein-coding genes…”

“The PhyloCSF tracks differ from other conservation browser tracks such as phyloP (Pollard et al. 2010) and phastCons (Siepel et al. 2005) in that the PhyloCSF tracks represent a signal of constraint specifically for protein-coding function, whereas the signal represented by other tracks is…”

“Meanwhile, ribosome profiling (RP) circumvents the experimental challenges in working with proteins, capturing sequence from the region of an RNA molecule that is attached to a ribosome (Ingolia et al. 2009).”

“At present, the two main human gene annotation projects, Ensembl/GEN-CODE (Harrow et al. 2012; Frankish et al. 2018; Zerbino et al. 2018) (henceforth GENCODE) and RefSeq (O ’ Leary et al. 2016), as well as the UniProt protein resource (The UniProt Consortium 2019), disagree on the number of human…”

“…used by several projects seeking to provide catalogs of protein function, including neXtProt (via UniProt) and the Human Protein Atlas (Gaudet et al. 2017; Thul et al. 2017); we anticipate that such resources will in due course provide valuable insights into these genes based on experimental data.”

“In fact, our 5′ UTR–associated examples include three cases in which protein existence has been confirmed by others through laboratory work: within the 5′ UTRs of MIEF1 (Rathore et al. 2018), MKKS (Akimoto et al. 2013), and RAB34 (Zougman et al. 2011).”

“The PhyloCSF tracks differ from other conservation browser tracks such as phyloP (Pollard et al. 2010) and phastCons (Siepel et al. 2005) in that the PhyloCSF tracks represent a signal of constraint specifically for protein-coding function, whereas the signal represented by other tracks is…”

“Meanwhile, we and others have previously found novel CDSs using mass spectrometry (Slavoff et al. 2012; Kim et al. 2014; Wilhelm et al. 2014; Weisser et al. 2016; Wright et al. 2016).”

“These data have been used to suggest the biological relevance of thousands of currently unannotated vertebrate open reading frames (ORFs) (Ingolia et al. 2011; Bazzini et al. 2014; Fields et al. 2015; Mackowiak et al. 9These authors contributed equally to this work.”

“…loci that we recently reported (Wright et al. 2016) based on a concurrent reanalysis of large “ draft proteome ” peptide data sets (Kim et al. 2014; Wilhelm et al. 2014) and all six loci from our analysis of testis data from the Chromosome-Centric Human Proteome Project (Supplemental Data S1, S2;…”

“Finally, we also recognize that certain reclassifications of lncRNAs as protein-coding genes would seem to contradict the findings of previous studies; this includes TUNAR (Lin et al. 2014) and TINCR (Kretz et al. 2013), both of which have ascribed noncoding functions.”

“However, it remains unclear what proportion of these are regulatory ORFs that do not produce functional peptides and instead compete with the canonical CDS for ribosome binding, or else are simply stochastic interactions (Calvo et al. 2009; Bazzini et al. 2014; Johnstone et al. 2016).”

“We found thatan addition-alsevenofthesegenescurrentlyhavepeptidesupportaccordingto neXtProt/PeptideAtlas criteria (Supplemental Data S2), although these are less stringent and include samples from cancer cell lines.”

“…et al. 2016) has increased substantially in recent years, and although such data sets have been used to discover novel proteins in genomes including human (Slavoff et al. 2012), difficulties remain in the creation and interpretation of high-quality “ proteogenomic ” data sets (Nesvizhskii 2014).”

“2017), STRIT1 (Nelson et al. 2016), and MYMX (Bi et al.”

“…et al. 2016) has increased substantially in recent years, and although such data sets have been used to discover novel proteins in genomes including human (Slavoff et al. 2012), difficulties remain in the creation and interpretation of high-quality “ proteogenomic ” data sets (Nesvizhskii 2014).”

“…unclear to what extent ribosome attachment shows production of a functional protein, that is, one that makes a direct contribution to physiology (Bazzini et al. 2014), because ORFs can also undergo translation as part of gene regulation mechanisms, and a proportion of attachments could be…”

“Transcription start sites were annotated based on cap analysis of gene expression (CAGE) libraries generated by FANTOM (The FANTOM Consortium and the RIKEN PMI and CLST (DGT) 2014), and polyadenylation sites were identified using PolyA-seq data (Derti et al. 2012).”

“…sets to distinguish those with protein-coding potential without having to install and run PhyloCSF (Makarewich et al. 2018; Perry et al. 2018; Huang et al. 2019a,b; Kang et al. 2019; Lin et al. 2019; McCorkindale et al. 2019; van Heesch et al. 2019; Vignal and Eory 2019; Wang et al.2019).”

“Although we have found that such publications may report an excess of false-positive novel protein-coding genes (Uszczynska-Ratajczak et al. 2018), they are also likely to have underreported the set of true-positives awaiting discovery because they generally targeted existing transcript catalogs,…”

“2016), and three were experimentally characterized, namely SPAAR (Matsumoto et al. 2017), STRIT1 (Nelson et al.”

“Transposon overlaps were found by comparing novel CDS to RepeatMasker regions (Smit et al. 2013) obtained from the UCSC Genome Browser (Casper et al. 2017), excluding regions of repeat class Low_complexity and Simple_repeat.”

“These PCCRs could represent upstream alternative splicing events that have not yet been captured in transcript libraries or perhaps show the usage of non-ATG initiation codons (Kearse and Wilusz 2017).”

“SMIM38was reported as translated based on proteomics data (Ma et al. 2016), and three were experimentally characterized, namely SPAAR (Matsumoto et al. 2017), STRIT1 (Nelson et al. 2016), and MYMX (Bi et al. 2017).”

“This is important because vertebrate genomes are replete with pseudogenes (see below) (Pei et al. 2012; Sisu et al. 2014).”

“However, it remains unclear what proportion of these are regulatory ORFs that do not produce functional peptides and instead compete with the canonical CDS for ribosome binding, or else are simply stochastic interactions (Calvo et al. 2009; Bazzini et al. 2014; Johnstone et al. 2016).”

“Next, as previously noted, we found that the CDSs identified within the 5 ′ UTRs of MKKS and MIEF1 had already been reported in published studies (Akimoto et al. 2013; Andreev et al. 2015; Delcourt et al. 2018), although these findings had not propagated into any annotation catalogs.”

“Also, C12orf81 was independently included in the CHESS database based on an earlier GenBank prediction (Pertea et al. 2018).”

“2015), capture-seq Pacific Biosciences (PacBio) data (Lagarde et al. 2017) and a vast collection of publicly available short-read RNA-seq datasets as processed by the Intropolis project (Nellore et al.”

“Wefoundthat61ofthe144genesexistedbeforethisstudy in either the RefSeq or UniProt catalogs or were previously characterized as ORFs by Mackowiak et al. (2015) based on their usage of PhyloCSF.”

“…sets to distinguish those with protein-coding potential without having to install and run PhyloCSF (Makarewich et al. 2018; Perry et al. 2018; Huang et al. 2019a,b; Kang et al. 2019; Lin et al. 2019; McCorkindale et al. 2019; van Heesch et al. 2019; Vignal and Eory 2019; Wang et al.2019).”

“Previously, we developed Phylogenetic Codon Substitution Frequencies (PhyloCSF) to support CDS annotation based on multispecies genome alignments (Lin et al. 2011).”

“” Briefly, in addition to sequences from the GenBank repository, annotation was also supported by SLR-seq (Tilgner et al. 2015), cap-ture-seq Pacific Biosciences (PacBio) data (Lagarde et al. 2017), and a vast collection of publicly available short-read RNA-seq data sets as processed by the…”

“By “ novel ” we do not mean de novo, that is, arising from noncoding sequence (Schlötterer 2015); in fact, many of these sequences have known orthologs in other species or paralogs in the species under consideration.”

“These data have been used to suggest the biological relevance of thousands of cur-rentlyunannotatedvertebrateopenreadingframes(ORFs)(Ingolia et al. 2011; Bazzini et al. 2014; Fields et al. 2015; Mackowiak et al. 2015; Raj et al. 2016).”

“Finally, we used the SORFS.org database of ORFs under 100 aa predicted from a comprehensive set of RP studies (Olexiouk et al. 2017) to find evidence of translation for six of our 50 CDS matching this size criterion (Supplemental Data S2).”

“…that represents stochastic “ noise ” remains to be ascertained (Pertea et al. 2018; Wan and Larson 2018), and although it could be that only a minority of transcript isoforms are translated into mature proteins, this remains highly debated (Mudge et al. 2013; Blencowe 2017; Tress et al. 2017).”

“…of PhyloCSF and its prede-cessor, CSF, to add CDS annotation to genomes within the Schizo-saccharomyces (Lin et al. 2011) and Drosophila lineages (Lin et al. 2007; The modENCODE Consortium et al. 2010; Jungreis et al. 2011, 2016), and also to identify novel human and mouse protein-coding genes…”

“…researchers investigating novel transcript sets to distinguish those with protein-coding potential without having to install and run PhyloCSF (Makarewich et al. 2018; Perry et al. 2018; Huang et al. 2019a,b; Kang et al. 2019; Lin et al. 2019; McCorkindale et al. 2019; van Heesch et al. 2019;…”

“…lacks cDNA or EST support but whose intron is confidently supported by short read data in Intropolis (blue rectangle) mostly from a retinal study (Farkas et al. 2013), and whose TSS (P1) is supported by FANTOM5 CAGE data, limited to retinaandeye(datafrom ZENBU browser,…”

“Human Protein Atlas (HPA) RNA-seq and human and mouse FANTOM5 CAGE data show high transcription in gastrointestinal tissues.”

“SMIM38was reported as translated based on proteomics data (Ma et al. 2016), and three were experimentally characterized, namely SPAAR (Matsumoto et al.”

“…makes a direct contribution to physiology (Bazzini et al. 2014), because ORFs can also undergo translation as part of gene regulation mechanisms, and a proportion of attachments could be stochastic “ noise ” (Guttman et al. 2013; Jackson and Standart 2015; Johnstone et al. 2016; Raj et al. 2016).”

“Small CDSs are harder to identify in both manual and computational annotation pipelines (Mudge and Harrow 2016), and this problem is confounded by the fact that 28 of these 50 loci are single-exon genes.”

“Mouse ENCODE RNA-seq supports placenta and ovary expression only, and the mouse locus (in the guise of a ncRNA) had previously been identified as a target for the germ cell–specific transcription factor Figla (Joshi et al. 2007).”

“…sets to distinguish those with protein-coding potential without having to install and run PhyloCSF (Makarewich et al. 2018; Perry et al. 2018; Huang et al. 2019a,b; Kang et al. 2019; Lin et al. 2019; McCorkindale et al. 2019; van Heesch et al. 2019; Vignal and Eory 2019; Wang et al.2019).”

“In fact, our 5 ′ UTR – associated examples include three cases in which protein existence has been confirmed by others through laboratory work: within the 5 ′ UTRs of MIEF1 (Rathore et al. 2018), MKKS (Akimoto et al. 2013), and RAB34 (Zougman et al. 2011).”

“…from the GenBank repository, annotation was also supported by SLR-seq (Tilgner et al. 2015), cap-ture-seq Pacific Biosciences (PacBio) data (Lagarde et al. 2017), and a vast collection of publicly available short-read RNA-seq data sets as processed by the Intropolis project (Nellore et al. 2016).”

“This is important because vertebrate genomes are replete with pseudogenes (see below) (Pei et al. 2012; Sisu et al. 2014).”

“…novel transcript sets to distinguish those with protein-coding potential without having to install and run PhyloCSF (Makarewich et al. 2018; Perry et al. 2018; Huang et al. 2019a,b; Kang et al. 2019; Lin et al. 2019; McCorkindale et al. 2019; van Heesch et al. 2019; Vignal and Eory 2019;…”

“…the UniProt protein resource (The UniProt Consortium 2019), disagree on the number of human protein-coding genes (Abascal et al. 2018), and even when a gene is agreed to be protein coding, it is often unclear which transcripts within the locus are translated (Mudge et al. 2013; Tress et al. 2017).”

“The proportion of this complexity that represents stochastic “ noise ” remains to be ascertained (Pertea et al. 2018; Wan and Larson 2018), and although it could be that only a minority of transcript isoforms are translated into mature proteins, this remains highly debated (Mudge et al. 2013;…”

“The spectrawere searched against a sequence database composed of all GENCODE v27 CDS transcripts combined with PhyloCSF sequences; an equally sized decoy database generated using DecoyPyrat (Wright et al. 2016) was concatenated and used to control FDR. Peptides were filtered to a posterior error probability of less th…”

“One of these is the remnants of crescent , previously characterized in chicken (Pfeffer et al. 1997) and a recognized mammalian pseudogenization event (Kuraku and Kuratani 2011).”

“Next, as previously noted, we found that the CDSs identified within the 5 ′ UTRs of MKKS and MIEF1 had already been reported in published studies (Akimoto et al. 2013; Andreev et al. 2015; Delcourt et al. 2018), although these findings had not propagated into any annotation catalogs.”

“…the UniProt protein resource (The UniProt Consortium 2019), disagree on the number of human protein-coding genes (Abascal et al. 2018), and even when a gene is agreed to be protein coding, it is often unclear which transcripts within the locus are translated (Mudge et al. 2013; Tress et al. 2017).”

“as well as the UniProt protein resource (The UniProt Consortium 2019), disagree on the number of human protein-coding genes (Abascal et al. 2018), and even when a gene is agreed to be protein coding, it is often unclear which transcripts within the locus are translated (Mudge et al.”

“CDSs can also be identified through sequence conservation, and both the ratio of nonsynonymous to synonymous substitutions ( d N / d S ) and codon substitution frequencies can be diagnostic of protein evolution (Lin et al. 2008).”

“…sets to distinguish those with protein-coding potential without having to install and run PhyloCSF (Makarewich et al. 2018; Perry et al. 2018; Huang et al. 2019a,b; Kang et al. 2019; Lin et al. 2019; McCorkindale et al. 2019; van Heesch et al. 2019; Vignal and Eory 2019; Wang et al.2019).”

“…to genomes within the Schizo-saccharomyces (Lin et al. 2011) and Drosophila lineages (Lin et al. 2007; The modENCODE Consortium et al. 2010; Jungreis et al. 2011, 2016), and also to identify novel human and mouse protein-coding genes based on the alignment of 29 mammalian genomes…”

“Next, as previously noted, we found that the CDSs identified within the 5 ′ UTRs of MKKS and MIEF1 had already been reported in published studies (Akimoto et al. 2013; Andreev et al. 2015; Delcourt et al. 2018), although these findings had not propagated into any annotation catalogs.”

“…makes a direct contribution to physiology (Bazzini et al. 2014), because ORFs can also undergo translation as part of gene regulation mechanisms, and a proportion of attachments could be stochastic “ noise ” (Guttman et al. 2013; Jackson and Standart 2015; Johnstone et al. 2016; Raj et al. 2016).”

“UTR-associatedORFs are extensively detected in vertebrate RP studies (Ingolia et al. 2009, 2011).”

“This gene has been implicated in a wide range of diseases, including cancer, hearing loss, liver disease, and immune disorders (González-Mariscal et al. 2017).”

“One of these is the remnants of crescent , previously characterized in chicken (Pfeffer et al. 1997) and a recognized mammalian pseudogenization event (Kuraku and Kuratani 2011).”

“…humanCDS that are apparently not transcribed in higherprimates despite showing vertebrate conservation, and yet, there is published evidence that the human locus produces a functional protein; we infer this must be a truncated molecule (Decarreau et al. 2017).”

“…sets to distinguish those with protein-coding potential without having to install and run PhyloCSF (Makarewich et al. 2018; Perry et al. 2018; Huang et al. 2019a,b; Kang et al. 2019; Lin et al. 2019; McCorkindale et al. 2019; van Heesch et al. 2019; Vignal and Eory 2019; Wang et al.2019).”

“The PCCR clusters analyzed and the resulting annotations are reported in Supplemental Data S1, and detailed information about each of the PCCRs in these clusters is reported in Supplemental Data S6.”

“A similar effort for the chicken genome is already underway (Vignal and Eory 2019).”

“The PhyloCSF pipeline has also been run on the each new release of the human and mouse genome annotations to facilitate the discovery of additional novel coding genes, novel pseudogenes, and novel coding sequence (21).”

“Numbers of newly named protein-coding genes remain comparatively low (28 in the last 2 years) because our protein-coding dataset is relatively complete, but novel approaches to detecting proteincoding genes continue to yield new gene models; for example, we worked closely with GENCODE annotators to name novel protein-…”

“Chlamydomonas was recovered with the expected branching order (Pröschold et al. 2018), as was the monophyly and expected topology of the TGV clade (Nakada et al. 2019).”

“37 Notably, 50 of these newly open reading frames (sORFs) in the untranslated regions (red).”

“Indeed, it has not been included among new coding ORFs identified with this approach [200].”

“…miniprotein prediction Following the Human Genome Project, miniprotein discovery and protein-coding annotation relied heavily on computational methods (Figure 1) (Wang et al., 2004), particularly evolutionary-based alignment techniques (Lin et al., 2011; Mudge et al., 2019; Wang et al., 2013).”

“However, this is unlikely to impact thedriverlncRNAspresentedhereonaccountofaggressive fi lteringof input annotations and manually checking of all driver predictions using PhyloCSF predictions 65 .”

“EV1C) (Mudge et al, 2019).”

“We computed PhyloCSF protein-coding scores for every three-nucleotide interval, in all three reading frames of SARS-CoV-2, smoothed using a hidden Markov model, and created tracks for the UCSC Genome Browser quantifying protein-coding evolutionary signatures along the genome, as we previously did for the human genome…”

“Phy-loCSF analysis of CDKN1A gene confirmed the positive selection ( average score 4.89 ) for CDKN1A full-length mRNA ORF ( Figure 5 A, 63 , 64 ) , whereas average PhyloCSF score for SPUD-ALE in the same frame as uORF ATG was -4.43.”

“As examples of its recent application, PhyloCSF scores (available as genomic tracks) were used to identify 133 novel protein coding genes in human [5] and to investigate the conservation of the newly emerged ORF10 in SARSCoV2 [2].”

“Correct read alignment across splice junctions of the raw data was performed by Gencode v25 gtf [83].”

“…PhyloCSF datasets for several species led to the detection of a set of short human proteins conserved within the mammalian order, several of which have been experimentally characterized (Anderson et al., 2015; Cloutier et al., 2020; Mudge et al., 2019; Pauli et al., 2014; Rathore et al., 2018).”

“The coding sequence for CCDC201 was previously missed due to a lack of spliced cDNA or expressed sequence tag evidence because its expression is restricted to female tissues 21 .”

“…which smORF_type was not annotated or ncRNA) to the annotations in the latest GENCODE release, version 38, which contains many protein-coding regions and pseudogenes that were not present in version 19, including several dozen added as a result of the Mackowiak et al. analysis (Mudge et al. 2019).”

“Characterizing smORFs that encode functional proteins among the millions of putative smORFs has remained challenging(94) .”

“No-tably, genome-wide association studies (GWAS) have revealed that some of these sORFs overlap disease-associated regions, implying potential correlation of sORF function with diseases [ 33 ].”

“…Codon Substitution Frequencies that considers not only the conservation of the coding sequence but also the high frequencies of synonymous codon and conservative amino acid substitutions as well as the low frequencies of other missense and nonsense substitutions (PhyloCSF; Mudge et al, 2019).”

“Recent advances in genomics, proteomics, and bioinformatics have allowed researchers to empirically define protein coding regions within genomes with better precision [2–4].”

“To assess coding potential, PhyloCSF scores were extracted for ORFs ≥ 10 codons from the Broad Institute dm6 PhyloCSF track hub, which provides strand-and frame-specific scores for the dm6 assembly [ 40 ].”

“From a biocuration perspective, there has been relatively little development of the “ conservome ” for coding sequences, that is, large-scale screens of conserved sequence elements with characteristic signatures of evolutionarily conserved coding sequences (Mudge et al. 2019).”

“Recently, PhyloCSF has been used for a genome-wide scan to discover new human protein-coding genes, a ”needle-in-a-haystack problem” [Mudge et al., 2019].”