SMBE 2026 Flash Talk

We downloaded bulk RNA-sequencing reads for non-pathological liver samples from 110 mammal species from the NCBI sequence read archive (SRA). We mapped these reads to RefSeq reference transcriptomes using Salmon to estimate transcript-level abundance (Patro et al. 2017). We identified orthogroups for the longest transcript of each gene for each species using OrthoFinder (Emms & Kelly 2019), and we subset these to single-copy orthogroups using OrthoSNAP (Steenwyk et al. 2022).  Shown on the right are normalized gene expression values (log2(TMM + 1)) for ALDH1A1 for the species in the tree on the left, derived from the maximum clade credibility tree of Upham et al. (2019). Colors represent diets derived from MammalDIET2 (Gainsbury et al. 2018) or EltonTraits (Wilman et al. 2014).

We implemented phylogenetic generalized least squares (PGLS) models using the caper R package to test for associations between gene expression and dietary traits while accounting for shared evolutionary history among species (Orme et al. 2013). We used the maximum clade credibility phylogeny from Upham et al. (2019) and diet for extant species defined as described to the left. Shown are normalized gene expression levels (log2(TMM + 1) for four of the 26 genes found to be significant at FDR = 0.05.

Shown above are the relative evolutionary rates of the element in the intersection of the Venn diagram to the right for all branches included in the RERconverge analysis of either diet or normalized gene expression of ALDH1A1, plotted against inferred change in log2(TMM + 1). Colors represent inferred dietary state of the downstream node or tip of each branch, derived from an ancestral state reconstruction using the method described in Redlich et al. (2023). Gray points represent branches for which diet is unknown or that were excluded from the analysis due to filtering.

We extracted all 300 bp sliding windows (sliding by 100 bp) within a region 1 Mb upstream and downstream of the human coordinates of ALDH1A1 from the human-referenced MAF of the VGP phase I whole-genome Cactus alignment (Formenti et al. 2026). We filtered these windows to retain ones with at least 50% of sequence aligned for at least 50% of the 166 mammals in the alignment, resulting in 7,793 windows. We then inferred estimated topology-constrained maximum likelihood trees for each element using phangorn with a GTR substitution model, based on the tree topology of the maximum clade credibility phylogeny from Upham et al. (2019). We tested for association of relative evolutionary rates (RER) with diet, as defined above, and with change in normalized gene expression of ALDH1A1, using RERConverge (Kowalczyk et al. 2019, Redlich et al. 2024). Shown above are the non-overlapping windows for which the raw p-value was less than 0.01 for association between RER and diet (orange) and/or between RER and normalized liver gene expression (log2(TMM + 1)) (blue).