Aster Plant

The rooted species tree is required in order to identify the correct out-group in each orthogroup tree, as correct gene tree rooting is critical for the orthology assessment from that tree [22 (link)]. Since orthogroups can potentially contain any subset of the species in the analysis, it is not sufficient to simply know the out-group for the complete species set. Instead, the complete rooted species tree is required. If the user knows the rooted species tree for the set of species being analyzed, then it is recommended to specify this tree manually at the command line to remove the possibility of species tree inference error. Such a tree can be provided as a Newick format text file. In the event that a species tree is not provided (or not known), then OrthoFinder automatically infers it.
Sets of one-to-one orthologs that are present in all species are often used for species tree inference; however, in real-world large-scale analyses, these can be rare [33 ]. A new algorithm, Species Tree from All Genes (STAG), was developed to allow species tree inference even for species sets with few or no complete sets of one-to-one orthologs present in all species [33 ]. Without this algorithm, species tree inference could fail if there were no sets of one-to-one orthologs present in all species. STAG infers the species tree using the most closely related genes within single-copy or multi-copy orthogroups. In benchmark tests, STAG [24 (link)] had higher accuracy than other leading methods for species tree inference, including maximum likelihood species tree inference from concatenated alignments of protein sequences, ASTRAL [38 (link)] and NJst [39 (link)].
The Species Tree Root Inference from Duplication Events (STRIDE) algorithm [22 (link)] is used to root the species tree in OrthoFinder. STRIDE was developed to enable the rooting of the species tree using only information available in the set of gene trees. STRIDE does this by identifying the set of well-supported in-group gene duplication events in the complete set of unrooted orthogroup trees, and using these events to infer a probability distribution over an unrooted STAG species tree for the location of its root. Similarly to STAG, STRIDE has been shown to identify the correct root of the species tree in multiple large-scale molecular phylogenetic data sets spanning a wide range of time scales and taxonomic groups [22 (link)]. In some cases, it is possible that there could be few duplications within the gene trees, and so STRIDE will not be able to identify the root of the species tree, or will only be able to exclude the root from clades in which gene duplication events are observed. In this case, ortholog inference should still not be significantly impacted since the rooting of the gene tree only affects ortholog inference in cases where gene duplication events are present [22 (link)]. This makes the STRIDE approach particularly suited to gene tree rooting for ortholog inference.

ASTRAL-II runs in $O\left(\mathit{\text{nk}}|X{|}^{1.726}\right)$ and $O\left(n^{3}k|X{|}^{1.726}\right)$ , respectively, with and without polytomies in gene trees.
In ASTRAL-I, X is the set of all bipartitions observed in input gene trees. While sufficient for statistical consistency and often for accuracy, under some conditions, this set X is too restrictive. To address this shortcoming, ASTRAL-II [22 (link)] uses several heuristics (see Additional file 1: Appendix A) and further expands the set X. Even though ATRAL-II tries to limit |X|, it does not provide any guarantees as to how it grows with n and k. In the worst case, |X| can grow exponentially, and thus, ASTRAL-II does not guarantee polynomial running time. The relatively high accuracy of ASTRAL-II has been shown in several simulations [20 (link), 22 (link), 29 (link), 30 (link)] and it has been adopted by the community as one of the main methods used in phylogenomics. ASTRAL has the ability to compute branch lengths in coalescent units [2 (link)] and a measure of branch support called local posterior probability [31 (link)].

As with JPred3, JPred4 makes secondary structure and residue solvent accessibility predictions by the JNet algorithm (11 (link),31 (link)). However, in JPred4, the JNet 2.0 neural network-based predictor has been retrained to make JNet 2.3.1 by 7-fold cross-validation using one representative for each of the 1358 SCOPe/ASTRAL v.2.04 superfamily domain sequences (32 (link)). Multiple alignments for each sequence were built by PSI-BLAST (33 (link)) through searching UniRef90 v.2014_07 (34 (link)). In addition to retraining, the HMM building step in JNet was updated to HMMer 3 (35 (link)) and some improvements were made to the code to simplify management and future algorithmic developments. The final accuracy of JNet 2.3.1 was assessed in a blind test on 150 sequences from 150 superfamilies not used in training. The 150 superfamily sequences were selected to reproduce a similar distribution of secondary structure compositions as the training structures in order to avoid biasing the reported accuracy of the blind test results. On the blind test, the average secondary structure prediction Q₃ score increased to 82.0% from 81.5% for JNet v.2.0, and solvent accessibility prediction accuracy rose to 90.0, 83.6 and 78.1% from 88.9, 82.4 and 77.8% for JNet v.2.0 for each of >0, >5 and >25% relative solvent accessibility thresholds.

With the development of high-throughput sequencing technology, more and more nuclear genes of species are obtained for phylogenomic analyses. Phylogenetics inference through large-scale genes concatenated into a supermatrix has proven to be flawed, such as being prone to systematic errors (and artifacts) and leading to an inaccurate phylogenetic relationship (Philippe et al., 2017 (link)). To understand the evolutionary history of tea plant populations cultivated in Xinyang, we constructed their phylogeny from both coalescent and ML methods by using low-copy nuclear genes and SNPs, respectively.
For coalescent analyses with 1785 low-copy nuclear genes, 94 new assemblies of sampled C. sinensis and two genome data (CSA ‘Yunkang 10’ and CSS ‘Shuchazao’) (Xia et al., 2017 (link); Xia et al., 2020b (link)) were used for phylogeny construction. Amino acid sequences were aligned using MAFFT v7.487 (Katoh and Standley, 2013 (link)) with the “-auto” parameter. Poorly aligned regions were further trimmed using the trimAl v1.2 (Capella-Gutiérrez et al., 2009 (link)) with the “-automated1” parameter. Multiple amino acid sequence alignments were converted to nucleotide alignments by PAL2NAL (Suyama et al., 2006 (link)). Single-gene ML trees were reconstructed using IQ-TREE v2.1.4-beta (Nguyen et al., 2015 (link)) under the GTR+ G model with 1000 bootstrap replicates. The coalescent analysis was implemented by ASTRAL.5.7.8 (Zhang et al., 2018 (link)).
For ML analyses by concatenating SNPs, a total of 108 samples included the 94 newly sequenced transcriptomes, and the RNA-seq data of CSA ‘Yunkang 10,’ and CSS ‘Biyun,’ ‘Hangdan,’ ‘Tieguanyin,’ ‘Longjing43’ and ‘Shuchazao,’ (Xia et al., 2017 (link); Wang et al., 2020 (link); Xia et al., 2020b (link); Zhang et al., 2020c (link); Wang et al., 2021b (link)) and eight wild tea species (Supplementary Table S2). The SNP dataset was converted to a PHYLIP file using the Python script ‘vcf2phylip’ (https://github.com/edgardomortiz/vcf2phylip/, accessed June 2022). The ML phylogeny was also inferred using IQ-TREE (Nguyen et al., 2015 (link)). The optimum model was selected with the maximum Bayesian Information Criterion (BIC) scores estimated by ModelFinder (Kalyaanamoorthy et al., 2017 (link)) implemented in IQ-TREE. Principal component analysis (PCA) was performed using Plink v1.90b6.25 (Purcell et al., 2007 (link)).