XB-ART-57344Nucleic Acids Res 2020 May 21;489:e51. doi: 10.1093/nar/gkaa142.
Show Gene links Show Anatomy links
TimeMeter assesses temporal gene expression similarity and identifies differentially progressing genes.
Comparative time series transcriptome analysis is a powerful tool to study development, evolution, aging, disease progression and cancer prognosis. We develop TimeMeter, a statistical method and tool to assess temporal gene expression similarity, and identify differentially progressing genes where one pattern is more temporally advanced than the other. We apply TimeMeter to several datasets, and show that TimeMeter is capable of characterizing complicated temporal gene expression associations. Interestingly, we find: (i) the measurement of differential progression provides a novel feature in addition to pattern similarity that can characterize early developmental divergence between two species; (ii) genes exhibiting similar temporal patterns between human and mouse during neural differentiation are under strong negative (purifying) selection during evolution; (iii) analysis of genes with similar temporal patterns in mouse digit regeneration and axolotl blastema differentiation reveals common gene groups for appendage regeneration with potential implications in regenerative medicine.
PubMed ID: 32123905
PMC ID: PMC7229845
Article link: Nucleic Acids Res
Species referenced: Xenopus
Genes referenced: a1cf abhd2 c1ql1 comt epha8 hs3st3a1 ilk ogdh pdgfd plin2 slc4a10 sult1a1
GO keywords: forebrain neuron development
Article Images: [+] show captions
|Figure 1. Illustration of TimeMeter by simulated high density discrete data. (A–D) Time shift pattern. (E–H) Different speed of dynamical change pattern. (I–L) Mixed pattern: the query has a 2-fold faster in dynamical change compared to the reference for the first 50 days, but after the first 50 days, the query has the same speed of dynamical change with the reference. TimeMeter uses DTW to align gene expression values, and then prunes excessively duplicated start or end points aligned indices, and truncates corresponding time points (C, G and K; dashed lines). TimeMeter applies piecewise (segmented) regression to aligned time points (after truncation), and partitions them into separate segments if more than one pattern is detected, such as figure (I). A progression advance score (PAS) is calculated by aggregation of area difference in each segment and normalized by total aligned time length (after truncation) in query.|
|Figure 2. Simulation study of how the data noise and sampling density will affect P-values in TimeMeter. The query and the reference have a time shift pattern (simulated discrete time series data). (A) Increasing the noise level will decrease the power to detect the pattern associations. (B) A higher sampling density will increase the power to detect the pattern associations.|
|Figure 3. Comparison of axolotl and Xenopus during early embryonic development. TimeMeter detects 2493 genes with similar temporal patterns (STP) between these two species. (A) Examples of STP genes with different PAS. (B) PAS distribution of STP genes. (C) Correlation between axolotl developmental stages and aligned Xenopus stages of STP genes.|
|Figure 4. Differential progression genes (|PAS| > 4) between axolotl and Xenopus during early embryo development. (A) Enriched neural development related GO terms in Axolotl advanced genes. (B) Enriched muscle or smooth muscle related GO terms in Xenopus advanced genes. (C–F) Examples of Axolotl advanced neural development/maturation markers. (G–I) Examples of Xenopus advanced muscle or smooth muscle markers.|
|Figure 5. Comparison of TimeMeter and Barry et al. for detecting genes with similar temporal patterns (STP) between human ES (from day 0 to day 42) and mouse EpiS cells (from day 0 to day 21) during neural differentiation. (A) Overlap of STP genes detected by Barry et al. and TimeMeter. (B) TimeMeter significantly increases the specificity for detecting STP genes. Barry et al. detected STP genes are enriched in 24 development related GO terms (P.adj < 0.05) (black triangle). None of these development related GO terms is enriched (P.adj < 0.05) in Barry et al. only gene list. In contrast, 20 out of 24 development related GO terms showed noticeable increased statistical significance for 1260 STP genes which were also detected by TimeMeter. (C–E) Examples of STP genes detected by both TimeMeter and Barry et al. (F, G) Examples of STP genes which were detected only by Barry et al. but not by TimeMeter. (H, I) STP genes which were detected only by TimeMeter but not by Barry et al.|
|Figure 6. The increased specificity for detecting genes with similar temporal pattern (STP) of TimeMeter is not at the cost of losing sensitivity. There are 32 development related GO terms are enriched (P.adj < 0.05) in either TimeMeter or Barry et al. detected STP genes between human ES and mouse EpiS during neural differentiation. There are eight terms specifically enriched in TimeMeter detected STP genes (but not enriched in Barry et al. list) while there is only one term marginally enriched in Barry et al. list (but not enriched in TimeMeter list).|
|Figure 7. PAS distribution of genes with similar temporal patterns (STP) between human ES and mouse EpiS during neural differentiation.|
|Figire 8. Nonsynonymous and synonymous substitution rates for temporally similar and dissimilar genes between human ES and mouse EpiS during neural differentiation. (A) Nonsynonymous substitution rate (dN). (B) Synonymous substitution rate (dS). (C) dN/dS ratio.|
|Figure 9. Genes with similar time-order patterns (STP) during mouse limb regeneration and axolotl blastema differentiation. (A) Enriched GO terms. (B–G) Examples of STP genes.|
|Figure 8. Nonsynonymous and synonymous substitution rates for temporally similar and dissimilar genes between human ES and mouse EpiS during neural differentiation. (A) Nonsynonymous substitution rate (dN). (B) Synonymous substitution rate (dS). (C) dN/dS ratio.|
References [+] :
Aach, Aligning gene expression time series with time warping algorithms. 2001, Pubmed