Background Gene annotation in eukaryotes is a non-trivial task that requires meticulous analysis of accumulated transcript data. Challenges include transcriptionally active regions of the genome that contain overlapping genes, genes that produce numerous transcripts, transposable elements and numerous diverse sequence repeats. Currently available gene annotation software applications depend on pre-constructed full-length gene sequence assemblies which are not guaranteed to be error-free. The origins of these sequences are often uncertain, making it difficult to identify and rectify errors in them. This hinders the creation of an accurate and holistic representation of the transcriptomic landscape across multiple tissue types and experimental conditions. Therefore, to gauge the extent of diversity in gene structures, a comprehensive analysis of genome-wide expression data is imperative. Results We present FINDER, a fully automated computational tool that optimizes the entire process of annotating genes and transcript structures. Unlike current state-of-the-art pipelines, FINDER automates the RNA-Seq pre-processing step by working directly with raw sequence reads and optimizes gene prediction from BRAKER2 by supplementing these reads with associated proteins. The FINDER pipeline (1) reports transcripts and recognizes genes that are expressed under specific conditions, (2) generates all possible alternatively spliced transcripts from expressed RNA-Seq data, (3) analyzes read coverage patterns to modify existing transcript models and create new ones, and (4) scores genes as high- or low-confidence based on the available evidence across multiple datasets. We demonstrate the ability of FINDER to automatically annotate a diverse pool of genomes from eight species. Conclusions FINDER takes a completely automated approach to annotate genes directly from raw expression data. It is capable of processing eukaryotic genomes of all sizes and requires no manual supervision—ideal for bench researchers with limited experience in handling computational tools.

Proteins encoded by newly-emerged genes (‘orphan genes’) share no sequence similarity with proteins in any other species. They provide organisms with a reservoir of genetic elements to quickly respond to changing selection pressures. Here, we systematically assess the ability of five gene prediction pipelines to accurately predict genes in genomes according to phylostratal origin. BRAKER and MAKER are existing, popular ab initio tools that infer gene structures by machine learning. Direct Inference is an evidence-based pipeline we developed to predict gene structures from alignments of RNA-Seq data. The BIND pipeline integrates ab initio predictions of BRAKER and Direct inference; MIND combines Direct Inference and MAKER predictions. We use highly-curated Arabidopsis and yeast annotations as gold-standard benchmarks, and cross-validate in rice. Each pipeline under-predicts orphan genes (as few as 11 percent, under one prediction scenario). Increasing RNA-Seq diversity greatly improves prediction efficacy. The combined methods (BIND and MIND) yield best predictions overall, BIND identifying 68% of annotated orphan genes, 99% of ancient genes, and give the highest sensitivity score regardless dataset in Arabidopsis. We provide a light weight, flexible, reproducible, and well-documented solution to improve gene prediction.

More than 15 petabases of raw RNAseq data is now accessible through public repositories. Acquisition of other 'omics data types is expanding, though most lack a centralized archival repository. Data-reuse provides tremendous opportunity to extract new knowledge from existing experiments, and offers a unique opportunity for robust, multi-'omics analyses by merging metadata (information about experimental design, biological samples, protocols) and data from multiple experiments. We illustrate how predictive research can be accelerated by meta-analysis with a study of orphan (species-specific) genes. Computational predictions are critical to infer orphan function because their coding sequences provide very few clues. The metadata in public databases is often confusing; a test case with Zea mays mRNA seq data reveals a high proportion of missing, misleading or incomplete metadata. This metadata morass significantly diminishes the insight that can be extracted from these data. We provide tips for data submitters and users, including specific recommendations to improve metadata quality by more use of controlled vocabulary and by metadata reviews. Finally, we advocate for a unified, straightforward metadata submission and retrieval system.

Background: With every new genome that is sequenced, thousands of species-specific genes (orphans) are found, some originating from ultra-rapid mutations of existing genes, many others originating de novo from non-genic regions of the genome. If some of these genes survive across speciations, then extant organisms will contain a patchwork of genes whose ancestors first appeared at different times. Standard phylostratigraphy, the technique of partitioning genes by their age, is based solely on protein similarity algorithms. However, this approach relies on negative evidence ─ a failure to detect a homolog of a query gene. An alternative approach is to limit the search for homologs to syntenic regions. Then, genes can be positively identified as de novo orphans by tracing them to non-coding sequences in related species. Results: We have developed a synteny-based pipeline in the R framework. Fagin determines the genomic context of each query gene in a focal species compared to homologous sequence in target species. We tested the fagin pipeline on two focal species, Arabidopsis thaliana (plus four target species in Brassicaseae) and Saccharomyces cerevisiae (plus six target species in Saccharomyces). Using microsynteny maps, fagin classified the homology relationship of each query gene against each target genome into three main classes, and further subclasses: AAic (has a coding syntenic homolog), NTic (has a non-coding syntenic homolog), and Unknown (has no detected syntenic homolog). fagin inferred over half the "Unknown" A. thaliana query genes, and about 20% for S. cerevisiae, as lacking a syntenic homolog because of local indels or scrambled synteny. Conclusions: fagin augments standard phylostratigraphy, and extends synteny-based phylostratigraphy with an automated, customizable, and detailed contextual analysis. By comparing synteny-based phylostrata to standard phylostrata, fagin systematically identifies those orphans and lineage-specific genes that are well-supported to have originated de novo. Analyzing within-species genomes should distinguish orphan genes that may have originated through rapid divergence from de novo orphans. Fagin also delineates whether a gene has no syntenic homolog because of technical or biological reasons. These analyses indicate that some orphans may be associated with regions of high genomic perturbation.

The evolutionary rapid emergence of new genes gives rise to "orphan genes" that share no sequence homology to genes in closely related genomes. These genes provide organisms with a reservoir of genetic elements to quickly respond to changing selection pressures. Gene annotation pipelines that combine ab initio machine-learning with sequence homology-based searches are efficient in identifying basal genes with a long evolutionary history. However, their ability to identify orphan genes and other young genes has not been systematically evaluated. Here, we classify the phylostrata of curated Arabidopsis thaliana genes and use these to assess the ability of two of the most prevalent annotation pipelines, MAKER and BRAKER, to predict orphans and other young genes. MAKER predictions are highly dependent on the RNA-Seq evidence, predicting between 11% and 60% of the orphan-genes and 95% to 98% of basal-genes in the annotated genome of Arabidopsis. In contrast, BRAKER consistently predicts 33% of orphan-genes and 98% of basal-genes. A less used method to identify genes is by directly aligning RNA-Seq data to the genome sequence. We present a Findable, Accessible, Interoperable and Reusable (FAIR) approach, called BIND, that mitigates the under-prediction of orphan genes. BIND combines BRAKER predictions with direct evidence-based inference of transcripts based on RNA-Seq alignments to the genome. BIND increases the number and accuracy of orphan gene predictions, identifying 68% of Araport11-annotated orphan genes and 99% of the conserved genes.