In this study, the eight Cucurbitaceae plant species, watermelon, bottle gourd, chayote, cucumber, melon, pumpkin, snake gourd, wax gourd, were used to identify OGs, respectively. The Cucurbitaceae genomes and annotation information were downloaded from the CuGenDB database¹. The assembled unique transcripts (PUT) from plant mRNA sequences were downloaded from PlantGDB², other 125 plant genome predicted proteins were downloaded from Phytozome³. UniProtKB was downloaded from Uniprot⁴ and NR database were downloaded from NCBI⁵, respectively.

To predict the potential functions of OGs, we downloaded published RNA-seq data to obtain the gene expression levels. These data included different tissues or different abiotic stresses of melon, cucumber and watermelon. We downloaded the transcriptome data of watermelon from NCBI⁶ under project number accession PRJNA454040 (leaf imposed to drought stress), PRJNA770012 (root imposed to osmotic stress) and PRJNA422970 (leaf and root imposed to different levels of nitrogen). The other RNA-seq resources of Cucurbitaceae plant species included melon (PRJNA383830: root, leaf, male flower, female flower, and fruit) and cucumber (PRJNA80169: root, stem, leaf, male flower, and female flower; PRJNA307098: Flower opening process, including green bud, green-yellow bud, yellow bud, and flowering).

We used the comparative genomics to detect OGs in the eight species from Cucurbitaceae. The identification pipeline is shown in Figure 1. For example, we first performed a BLASTP on the watermelon protein sequences against the proteome data of the other seven collected Cucurbitaceae plants. It was discarded if the watermelon protein sequence has a significant BLASTP hit in other species with an E-value < 1e^–5. We then performed a homolog search against Plant-PUTs database, other published plant genome sequences, UniProtKB database and NR database with an E-value < 1e^–5, respectively. Ultimately, the genes that match to neither other databases are the OGs in watermelon (Zhang et al., 2007; Lin et al., 2010), and genes with at least one homolog are NOGs. The other seven Cucurbitaceae species were performed utilizing the same identification process.

To analyze the characteristics of OGs in Cucurbitaceae, we downloaded the whole genome information of eight Cucurbitaceae species (see text footnote 1). Then in-house python scripts were used to calculate protein length, GC content, exon length, and the number of exons per gene⁷. The isoelectric point of the proteins was calculated using the DAMBE7 software (Xia, 2018). We used the Wilcox rank sum test to estimate significant differences between OGs and NOGs in different groups. BUSCA (Bologna Unified Subcellular Component Annotator) was then used to predict the subcellular localizations of OGs with eukarya plants mode (Savojardo et al., 2018). The chromosomal localization information of OGs was extracted from the annotation file, and then mapping was performed with Mapgene2chrom⁸.

The study of how genes arise and the differentiation process is essential to explain the generation and evolution of new phenotypes and finally the inheritance of biodiversity (Long et al., 2003). According to previous studies, there are four main mechanisms that account for how OGs emerged, including gene duplication, gene overlap, transposable element (TE) exaptation and de novo origin (Wu et al., 2011; Wissler et al., 2013). Among the four mechanisms, gene duplication is thought to be the predominant mechanism of origin (Zhang, 2003). We used DupGen_finder.pl to identify OGs originating from gene duplication, which can identify five different duplication modes, including tandem duplication, whole genome duplication (WGD), dispersed duplication, transposon duplication, and proximal duplication (Qiao et al., 2019). First, we aligned the protein sequences within a genome by BLASTP with an E-value < 1e^–8. Next, DupGen_finder.pl was used to determine the model of gene duplication based on the detected homologous gene pairs. Synonymous sites (Ks) were computed using the Nei–Gojobori approach implemented in the python script synonymous_calc.py⁹. Finally, the universal mutation rate of 6.5 × 10^–9 was used to assess the time of gene duplication of OGs (Gaut et al., 1996). To identity overlapping gene models, we used OGs against CDS sequences of other Cucurbitaceae species to screen for homologous sequences covering at least 50% of the length of gene. To identify the OGs overlapped with TEs, we first used the software RepeatMasker to identify TEs in eight Cucurbitaceae species, composition of a Cucurbitaceae TEs dataset (Mendeley Data)¹⁰ and then the CDS sequences of the Cucurbitaceae OGs were used as queries to do BLASTN searches against the Cucurbitaceae TE sequences dataset with an E-value < 1e^–5. For de novo originated genes, orphan proteins sequences of the remaining were then searched with TBLASTN against the genomes of seven other Cucurbitaceae to identify orthologous non-coding sequences. The orthologous non-coding sequence was defined according to three indicators: (1) at least 60% sequence identity and covering at least 80% of the orthologous regions of target gene could be aligned; (2) the alignment significance E-value < 1e^–6; (3) the protein lengths of other Cucurbitaceae species that have premature translational termination should be shorter than 50% of the length of the candidate orphan proteins (Wu et al., 2011; Sun et al., 2015; Zhang et al., 2019).

To further explore the role of OGs in Cucurbitaceae, we calculated the expression levels of genes in different tissues and identified differentially expressed genes (DEGs) in different treatments on the basis of the collected transcriptome data. We filtered RNA-seq data using Trimmomatic. RSEM was used to compute FPKM (fragments per killobase of exon per million fragments mapped) values. DEGs analysis was carried out with the DESeq2 R package (Love et al., 2014). The significant differences in gene expression were determined using a false discovery rate (FDR) set at < 0.05 and | log₂FC| > 1 as cutoffs. Genes with FPKM value > 0.02 were assumed to have been expressed (Ma S. et al., 2020). Besides, PaGeFinder software with specificity measure (SPM) was used to identify the genes specifically expressed in a certain tissue (Pan et al., 2012), and it was determined as a specific gene in this tissue once the SPM value was ≥ 0.9.

The co-expression network construction was completed using the WGCNA software package of R software. First, the data was filtered to construct the expression matrix and the samples were clustered for analysis. The soft threshold β function in WGCNA was used to calculate the coefficient of similarity among all genes to create the adjacency matrix and construct the systematic clustering tree. The minimum number of genes per module was set to 30, and the initial co-expression module was identified using cutreeDynamic. Subsequently, the eigenvector value ME (Module eigengene) of each module is calculated using the moduleEigengene function. The correlation coefficient between module ME and sample features is quantified using the cor function and a heat map is drawn. Higher correlation coefficients indicate a higher correlation between the genes in the module and the sample features, and accordingly, higher relative gene expressions. Modules with high correlation coefficients with sample characteristics are selected as the tissue-specific module. The module membership (MM) and gene significance (GS) of ME in each tissue-specific module were calculated. If MM > 0.95 and GS > 0.85, the gene was decided as the central gene of the module. In addition, we also used Fuzzy c-means clustering to analyze the transcriptome time series data (Parker and Hall, 2014). KEGG pathway enrichment analysis based on the KEGG biology pathway database and mapping was performed on OmicShare¹¹, which is an online platform.

Based on previous studies, we designed a comprehensive, systematic computational pipeline to identify OGs in the Cucurbitaceae genomes (Figure 1; Doerks et al., 2002; Tay et al., 2009; Yang et al., 2013; Xu et al., 2015). For instance, in the watermelon, there were 22,596 annotated protein-coding genes, which were used for BLASTP with the Cucurbitaceae genome. The 1971 genes (DatabaseI, DBI) were kept for follow-up analysis. The retained genes were then against with 125 plant genomes, and 1,945 genes had no match (DBII). In the next comparison of these genes with 251 PlantGDB-assembled Unique Transcripts (PUTs) sequences, 1674 genes were found to be non-homologous (DBIII). A final step to remove further the impact of false positives on the analysis was to analyze the remaining genes in comparison to the UniProtKB and NR databases, resulting in 1,652 genes being left. The final leftover 1,652 genes were termed as OGs in the watermelon genome (Supplementary Table 1). Using the same pipeline, we identified OGs for each of the other seven Cucurbitaceae species’ genomes (Supplementary Table 1). OGs contents greatly varied among the eight Cucurbitaceae species, ranging from 1.63% in chayote and 16.55% in wax gourd (Supplementary Table 2). The proportions of OGs in the different species varied greatly, typically in the range of 5-15%. Previous studies of sweet orange (Xu et al., 2015), rice (Guo et al., 2007), wheat (Ma S. et al., 2020), and Populus trichocarpa (Guo, 2013) reported 3.54, 3.23, 1.4, and 14.32% orphans, respectively. In two close relative species, they also showed great differences, the proportion of OGs in Arabidopsis thaliana is 5.3%, while it is 12.3% in Arabidopsis lyrate (Guo, 2013). Part of this variation is due to the different evolutionary distance that exists between each focal species and its closest sequenced relatives (Wissler et al., 2013). The more genomes of reference species are decoded, the more annotation messages are available, and the accuracy of prediction may become higher. Another part variation may be due to real differences in evolutionary pressures (Arendsee et al., 2014).

Among all species, OGs have a shorter origination time, and whether they have distinctive features relative to NOGs is an intriguing question. To determine the differences that exist in Cucurbitaceae species, we performed an analysis and compared the gene structure of OGs and NOGs. Our results showed that OGs exhibit significantly shorter protein lengths in all eight species of Cucurbitaceae (Figure 2A). The average protein length distribution of the OGs and NOGs were 48-155, 348-436 amino acids, ranging from 2.78 times in snake gourd 7.25 times in melon (Supplementary Table 3, Wilcox rank sum test, P-value < 0.001). This is a finding that is consistent with the sequence characteristics that have been reported in primates (Toll-Riera et al., 2009), zebrafish (Yang et al., 2013), sweet orange (Xu et al., 2015), and Caenorhabditis elegans (Zhang et al., 2019). In-depth analysis of the structural components of the genes showed that the shorter protein length was mainly attributed to the fewer number of exons (Figure 2B and Supplementary Table 3, Wilcox rank sum test, P-value < 0.001). In the zebrafish, about 28% of the OGs have only one exon, compared to the NOGs with only one exon is 6% (Yang et al., 2013). About 36.87% of the genes in the maize were intronless genes, of these, about 16.67% were OGs (Yan et al., 2014). In primate species, OGs also contained fewer exons compared to NOGs (Toll-Riera et al., 2008). This suggests the prevalence of these two characteristics for OGs in all eukaryotes. We also analyzed the exon lengths of OGs and found it was significantly longer in snake gourd, while it was significantly shorter in other Cucurbitaceae species (Figure 2C and Supplementary Table 3, Wilcox rank sum test, P-value < 0.001). In sweet orange, the exon length of OGs was shorter than that of NOGs, but in wheat Caenorhabditis elegans, the exon lengths played an insignificant role (Xu et al., 2015; Li G. et al., 2019; Ma S. et al., 2020). The reason for this is the evolutionary process of short new genes to long old genes involving mainly the recruitment of alternative exons rather than the expansion of individual exon lengths (Neme and Tautz, 2013). The GC content of some species increases progressively across the phylostrata (Donoghue et al., 2011; Wissler et al., 2013), and the youngest gene in Arabidopsis thaliana also had a sharply separated from non-genetic ORFs (43% median GC content for OGs and 32% median GC content for NOGs) (Lin et al., 2010). However, this characteristic is not universal. In Strobilanthes cusia, there was no difference in GC content (Hu et al., 2021). In Aegiceras corniculatum and wheat, OGs were significantly less than that of NOGs (Ma et al., 2021). Among Cucurbitaceae species, they also exhibited heterogeneously. In chayote and snake gourd, the GC content of OGs was not different compared to NOGs, and in cucumber, melon and wax gourd, OGs were significantly higher, while in watermelon, bottle gourd and pumpkin, significantly lower (Figure 2D and Supplementary Table 3, Wilcox rank sum test, P-value < 0.001). Changes in the isoelectric point are important indicators of altered protein function (Khaldi and Shields, 2011). We found that the isoelectric points of OGs were significantly higher than NOGs in all species of Cucurbitaceae (Figure 2E and Supplementary Table 3, Wilcox rank sum test, P-value < 0.001), which may be associated with the fact that species have to adapt to variable environments (Andrade et al., 1998; Kiraga et al., 2007).

Predicting the subcellular localization is important for understanding the nature and function of proteins in cells and exploring the interactions between proteins (Emanuelsson et al., 2007; Chou and Shen, 2008). We performed subcellular localization analysis on eight Cucurbitaceae species, and the results showed that the localization was mainly on the nucleus (pumpkin: 30.86% ∼ snake gourd: 52.4%) and chloroplast (snake gourd: 20.96% ∼ bottle gourd: 32.64%) (Figure 3A and Supplementary Table 1). Except for pumpkin, which was mostly localized on chloroplast, the other seven species were mostly localized on the nucleus. In general, the subcellular localization of Cucurbitaceae species was essentially the same, with only subtle differences.

The distribution properties of OGs on chromosomes differ between species. In zebrafish, the distribution of OGs on chromosomes was heterogeneous, with a high proportion of OGs on some chromosomes and none on others (Yang et al., 2013). The OGs in the wheat genome were evenly distributed on the 21 chromosomes, with a higher density in the regions near the telomeres (Ma S. et al., 2020). OGs in the Arabidopsis thaliana genome were not grouped but uniformly spread across the genome among NOGs (Donoghue et al., 2011). For the analysis of OGs’ genomic distribution, we plotted the OGs over the chromosomes of Cucurbitaceae based on the available information from genome annotation (Supplementary Table 1). In watermelon, the proportion of OGs distributed on 11 chromosomes ranged from 6.33% to 8.30%, bottle gourd (3.35% ∼ 4.42%), chayote (1.65% ∼3.23%), cucumber (8.85% ∼9.85%), melon (5.59% ∼8.5%), pumpkin (5.79% ∼9.22%), snake gourd (1.30% ∼2.54%) and wax gourd (15.17% ∼17.79%). In the pumpkin, OGs were unevenly distributed across the 20 chromosomes, with the largest distribution in Chr1 and the least distribution in Chr19 (Figure 3B), with an approximately 1.59-fold difference. In the melon, OGs were unevenly distributed across the 12 chromosomes, with the largest distribution in Chr10 and the least distribution in Chr7 (Figure 3B), with an approximately 1.52-fold difference. Compared to pumpkin and melon, the spread of OGs on chromosomes was reasonably uniform in other Cucurbitaceae species. Besides, the distribution density of OGs was higher near the telomeres, and the distribution was relatively balanced on the chromosomes apart from the aggregation phenomenon in some chromosomal regions (Supplementary Figure 1).

The evolutionary origin of OGs is a microevolutionary process in which the structure of a gene arises from the mutation of a germ cell gene, and the study of how genes arise and differentiate is essential to explain the generation and succession of novel phenotypes and eventually biodiversity (Tautz and Domazet-Lošo, 2011; Long et al., 2013). OGs are typically generated by multiple combinations of mechanisms of origin, and gene duplication is thought to be a major driving force of the production of OGs, for example, in Arabidopsis thaliana, rice, Drosophila, and primates (Guo et al., 2007; Toll-Riera et al., 2009; Donoghue et al., 2011). OGs originating from gene duplication were greatly varied among the eight Cucurbitaceae species, ranging from 50 (2%) in pumpkin 196 (37.05%) in snake gourd (Figure 4A and Supplementary Table 2). Most Cucurbitaceae species have a dispersed gene duplication type, except for pumpkin where the most dominant gene duplication type is whole genome duplication. We further estimated the duplication time of OGs in melon and wax gourd. We found that the duplication time of OGs was about 6.15∼15.38 MYA in melon, coinciding with the differentiation time of melon and cucumber (10 MYA), and maybe related to its correlated biologically relevant characters (Figure 4B; Garcia-Mas et al., 2012). In wax gourd, it was about 30.77∼37.26 MAY. This timing coincides with tribe Benincaseae, which was estimated to be distinct from the tribe Momordiceae, containing bitter gourd (36.1 MYA, Figure 4C; Xie et al., 2019). In addition, we also analyzed OGs originating from gene overlap, TE exaptation, and de novo models in Cucurbitaceae species (Supplementary Table 1). The results showed a large variation, and for the gene overlap originated OGs, ranging from 12 (1.91%) in chayote 103 (11.84%) in bottle gourd. For the TE exaptation originated OGs, ranging from 24 (1.04%) in pumpkin 129 (24.39%) in snake gourd. For the de novo originated OGs, ranging from 3 (0.57%) in snake gourd 106 (4.63%) in melon (Supplementary Table 2). Future studies will be needed to determine whether these OGs with different models of origin are functional and to reveal their relevance in the adaptive evolution and generation of new traits of agronomic importance of Cucurbitaceae species.

OGs are often functionally enigmatic due to the lack of homology and functional structural information. To reveal the potential biological functions of these OGs, we investigated the gene expression patterns based on transcriptome profiling data. In this study, we performed an analysis of the OGs expression based on previously published RNA-seq data in (1) five tissues in melon (root, leaf, male flower, female flower, and fruit); (2) six tissues in cucumber (root, stem, leaf, male flower, female flower, and ovary). Transcriptional data showed evidence of expression of 1,383 (60.47%) OGs and 24,358 (87.96%) NOGs in melon, and 937 (37.12%) OGs and 20,292 (96.82%) NOGs in cucumber. Further studies found, in melon and cucumber (Supplementary Table 4), there were 636 (27.81%) and 165 (17.61%) OGs, and 4,935 (17.82%) and 2,162 (10.65%) NOGs showed tissue-specific expression, respectively. It is evident that OGs were less abundantly expressed but had a higher tissue-specific expression, which is in agreement with the expression patterns of OGs observed in other species (Long et al., 2003; Wu et al., 2011) and these tissue-specific expressed genes may be associated with specific phenotypes or specific physiological processes. Interestingly, the largest proportion of these OGs was significantly more tissue-specific expressed in flower than in other tissues (melon: 69.97% and cucumber: 40.61%) (Supplementary Table 4), and more detectable expression was found in male flower (melon: 35.38% and cucumber: 21.82%) (Supplementary Table 4). Many studies have shown that OGs, or young genes as some researchers call them, are more inclined to be expressed in the male organs (McCarrey and Thomas, 1987; Yang et al., 2009; Cui et al., 2015; Ruiz-Orera et al., 2015).

Using co-expressed genes to predict the function of OGs is an effective method (Li G. et al., 2019; Zhang et al., 2019; Ma et al., 2021). To investigate the underlying functions of OGs in melon, we used WGCNA analysis to identify 12 co-expressed gene modules (Figure 5A). After screening, 1017 genes were identified in MEgreen module (tissue-specific expression in male flower), including 18 OGs (Supplementary Table 5 and Figure 5B). Functional annotation showed that co-expressed genes of OGs are frequently involved in carbohydrate metabolism (Figure 5C). Carbohydrates play a key role in the development of male gametophytes, providing nutrition for normal growth and possibly acting as signaling molecules to influence development in the process (Clément and Audran, 1995). Many male sterile lines have been revealed to be involved in disorders of carbohydrate metabolism (Dorion et al., 1996; Zhang et al., 2010; Min et al., 2013). In addition, KEGG analysis showed the co-expressed genes of OGs are significantly enriched (P-value < 0.05) in pentose and glucuronate interconversions, cutin, suberine and wax biosynthesis, starch and sucrose metabolism, ascorbate and aldarate metabolism, phenylpropanoid biosynthesis, and anthocyanin biosynthesis (Figure 5D). Transcriptome analysis of the flower organ of male-sterile and fertile plants of Allium sativum revealed a certain number of differential genes related to anthocyanin biosynthesis (Shemesh-Mayer et al., 2015). The sucrose transporter gene CsSUT1 of cucumber is expressed in the male flower, and down-regulation of CsSUT1-RNA interference (RNAi) expression can induce male sterility, a process that primarily affects starch and sucrose metabolism, and pentose and glucuronate interconversions (Sun et al., 2019). In addition, the anthocyanin biosynthesis, phenylpropanoid biosynthesis, and ascorbate and aldarate metabolism are engaged in the scavenging of reactive oxygen species in plants under adversity stress (Xu L. et al., 2017; Sharma et al., 2019). In the process of plant growth and development, due to nucleoplasmic genetic disharmony and various external adverse environments, excess reactive oxygen is produced, which leads to abnormal function of the cell membrane system and eventually causes male sterility (Fridovich, 1978; Li et al., 2004). The results of various studies shows that the potential functions of OGs specifically expressed in the male flower of melon are closely related to male sterility.

To investigate the role of OGs on flower development, we analyzed transcriptome data from four different flowering stages of cucumber, including green bud (1 day), green-yellow bud (3 days), yellow bud (4 days), and flowering (5 days). Compared with the control group (green bud), a total of 10,248 significantly differentially expressed genes (DEGs) were identified, including 182 OGs. Fuzzy c-means clustering analysis of all DEGs (including OGs and NOGs) was further divided into six Clusters of gene co-expression patterns (Figure 6A). Genes with memberships > 0.6 in the Cluster were screened for subsequent functional enrichment analysis. There were a total of 385 DEGs in Cluster 1 (decreasing), including 2 OGs, and 381 DEGs in Cluster 5 (increasing), including 7 OGs (Figure 6B). KEGG enrichment results for co-expressed genes of OGs showed that DNA replication, Porphyrin and chlorophyll metabolism, and flavone and flavonol biosynthesis were significantly enriched in Cluster 1. Peroxisome, endocytosis, and tyrosine metabolism were significantly enriched in Cluster 3 (P-value < 0.05, Figure 6C). Peroxisomes are small cellular organelles that produce a variety of metabolites and are essential in modulating plant growth and development (Xu S. et al., 2017; Kao et al., 2018). OsPEX5 encodes a peroxisome targeting sequence 1 (PTS1) receptor protein, whereas the mutants can cause abnormal rice spikelet morphology (You et al., 2019).

Evidence suggests that OGs are recruited into roles that regulate responses to changing environments (Gollery et al., 2006; Luhua et al., 2008). To probe the possible association between OGs and environmental adaptation in watermelon, we reanalyzed the expression of OGs in (1) suspension cells under osmotic stress (0, 2, and 4 h); (2) leaves under drought stress (4 days and 8 days); (3) roots and leaves under nitrogen (N) (LowN: 0.2 mM and HighN: 9 mM). In osmotic stress, compared with the control group (0 h), we identified 8 (up-regulated: 4, down-regulated: 4) and 3 down-regulated OGs in 2 h vs. 0h and 4 h vs. 0h were differentially expressed, respectively. A total of 9 OGs overlapped that were osmotic responsive (Supplementary Table 6). Fuzzy c-mean clustering analysis showed that all DEGs were classified into four Clusters (Figure 7A). The genes in the clusters with memberships > 0.6 were filtered out and subjected to KEGG enrichment analysis. There were a total of 1,049 DEGs in Cluster 1, including 5 OGs, and 423 DEGs in Cluster 3, including 2 OGs (Figure 7B). KEGG enrichment results for co-expressed genes of OGs showed that carbon fixation in photosynthetic organisms, Glycolysis/Gluconeogenesis, monoterpenoid biosynthesis, and zeatin biosynthesis pathway were significantly enriched in Cluster 1. Spliceosome, phosphatidylinositol signaling system, nitrogen metabolism, and plant hormone signal transduction were significantly enriched in Cluster 3 (P-value < 0.05, Figure 7C). In drought stress, a total of 4,090 DEGs were identified, of which 37 were OGs (Supplementary Table 6). A WGCNA for all DEGs associated with drought led to the further identification of two essential co-expression modules MEblack (drought treatment for 4 days) and MEturquoise (drought treatment for 8 days) (P-value < 0.05, Supplementary Figure 2), where 7 OGs were participating and enriched in GO terms, consisting ‘response to water’, ‘response to high light intensity’ and ‘response to hydrogen peroxide’ (P-value < 0.05, Supplementary Table 7). These correlations suggest a vehicle to infer the functions of OGs and indicate the incorporation of certain OGs into conserved stress networks as an underlying mechanism for watermelon to adapt to drought stress. In addition, we also identified three (Cla97C02G028900, Cla97C05G099800 and Cla97C09G167820) and two OGs (Cla97C06G126600 and Cla97C09G171550) that were differentially expressed in leaves and roots under nitrogen stress, respectively (Supplementary Table 6). Clearly, among the three abiotic stresses, OGs were more prominent under the drought condition.

The fruits of melons all undergo a process of small to large size, and wax gourd have larger fruits compared to other species in the Cucurbitaceae family. Xie et al. (2019) combined population genetics and linkage mapping to pinpoint 3,939 genes that may have been selected as part of domestication and improvement, with some possibly responsible for the large fruit size of wax gourd (Xie et al., 2019). Interestingly, by screening, we found 520 OGs (Supplementary Table 8), which represent about 13.2% (520/3939) of all genes domestication and improvement sweeps and about 11.44% (520/4546) of all OGs in the wax gourd genome.

In conclusion, we found that OGs played a crucial role in several aspects, such as male sterility, environmental adaptation and crop domestication. Our results provide the molecular basis and research direction for some important research mechanisms in Cucurbitaceae, but the specific functions of OGs need further experimental validation.