Figure 1.
Description of isoform multiplicity co-occurrence (IMco).
The figure illustrates how IMco is computed. We start from a set of human-mouse orthologs for which isoform annotations are available in ENSEMBL (or VEGA), as shown in (A). We will say that one gene has isoform multiplicity (IM) when it has more than one isoform. The scheme shows different instances that cover all possible IM combinations for human-mouse ortholog pairs: both orthologs have IM, only one has IM, and no ortholog has IM. (B) IMco will correspond to the fraction of cases for which both orthologs have IM; it is computed as shown in the table.
Figure 2.
Relationship between isoform multiplicity co-occurrence (IMco) and protein divergence (PD).
In (A), (B) and (C) we plot IMco (the fraction of (IM, IM) ortholog pairs, see Figure 1 and Materials and Methods) as a function of PD (measured using the percentage of sequence identity). Black dots are used to display the raw, unprocessed data; a blue line is used for the smoothed data and in grey we show the envelope (see Results section). In both cases we can see that there is a monotonically increasing relationship between IMco and PD (see text). Outliers from this trend define two lines, one at 0 and the other at 1; these outliers result from IMco estimates obtained with less than 5 observations. They essentially disappear when eliminating these poorer estimates (Figure S1). (A) and (B) where obtained with a Uniprot/SwissProt/RefSeq-based sequence dataset and Ensembl and VEGA isoform annotations, respectively; (C) was obtained using Ensembl data for both genes and transcripts. The monotonic trend is comparable in the three figures, although in (C) the curve shows a slight shift towards lower IMco values resulting from lower amount of genes with IM.
Table 1.
Statistics of the datasets used in this work: number of ortholog pairs and percentage of genes with IM (the latter refers only to the model species, not to human).
Figure 3.
The contribution of species-specific isoform multiplicity (IM) to isoform multiplicity co-occurrence (IMco).
Here we compare IMco with P(IMH|x)⋅P(IMM|x), the product of species-specific IM and a term of IMco, as shown in Equation 2 (see Materials and Methods). In (A) we show the raw data representation: we can observe an important overlap between both data clouds, as well as a similar monotonic trend, something confirmed in (B) where we show the smoothed data. In (A) we used black and grey for IMco and P(IMH|x)⋅P(IMM|x), respectively; in (B) we used dark and light blue, respectively.
Figure 4.
Species-specific isoform mutiplicity (IM) vs. protein divergence (PD).
Here we show the relationship between species-specific IM and PD, for both human and mouse genes. Black and grey dots are used for human and mouse, respectively. We observe the same monotonic trend for both species, a result that provides a simple explanation for the also monotonic behavior of P(IMH|x)⋅P(IMM|x) (see Figure 3), the product of species-specific IM. In addition, this result is an important intermediate step that will allow us to trace back the result in Figure 2 to a simple gene-level property (see text and Figure 6): the number of exons of the longest gene isoform. The continuous lines represent the smoothed version of the raw data (yellow and red for human and mouse, respectively; grey for the envelope) and lead to the same interpretation.
Figure 5.
The relationship between isoform multiplicity (IM) and number of exons of the largest isoform.
The figure illustrates the relationship between these two properties for both human and mouse genes (A), and for these and other species (B). (A) and (B) differ in the data origin: (A) was obtained using UniProt/SwissProt/RefSeq sequences and Ensembl/VEGA transcript annotations; (B) was obtained using only Ensembl data. In general, we see the same trend: an increasing monotonic relationship which approaches 1 asymptotically, indicating that the larger the number of exons of the gene, the larger the number of isoforms of this gene. The fluctuations observed are due to a combination of factors: irregular isoform annotations (e.g. fruit fly), or low sample effects (particularly, for number of exons>30). Chimpanzee is an exception due to a very low percentage of transcript annotations (only 5% of the genes had multiple isoforms in the version of Ensembl used).
Figure 6.
Number of exons vs. protein divergence (PD).
Here we show the relationship between number of exons of the largest isoform and PD, for both human and mouse genes. Black and grey dots are used for human and mouse, respectively. We observe the same monotonically increasing trend for both species, which provides a natural explanation for the behavior of species-specific IM seen in Figure 4. The smoothed version is shown with a continuous line (yellow and red for human and mouse, respectively; grey for the envelope). The relevance of this result is that we have identified a simple gene property contributing to the relationship between IM co-occurrence, IMco, and PD shown in Figure 2. (A) was obtained using UniProt/SwissProt/RefSeq sequences and Ensembl transcript annotations; (B) was obtained using only Ensembl data.
Figure 7.
Number of exons is a component of isoform multiplicity co-occurrence (IMco).
We display (black dots) the relationship of IMco vs. the product of human and mouse number of exons. We chose the product because it is a priori related to IMco through Equation (2) and Figure 5. The result shows the existence of a monotonically increasing relationship. Shown in blue is the smoothed version of the raw data, and the envelope in grey.
Figure 8.
Number of exons vs. protein divergence (PD) in other species.
This figure is equivalent to Figure 6, but in this case we show the results for six other species. In each plot we represent the data for human (black dots; smoothed version in yellow) and the other species (grey dots; smoothed version in red); the envelopes of the smoothed versions are shown in grey. From left to right and top to bottom we have the results for: chimpanzee (rho = 0.3 and 0.4, p-value∼10−10 and ∼10−17, for human and chimpanzee, respectively), cow (rho = 0.6 and 0.6, p-value∼10−59 and ∼10−66), rat (rho = 0.4 and 0.6, p-value∼10−25 and ∼10−66), chicken (rho = 0.3 and 0.4, p-value∼10−21 and ∼10−27), zebrafish (rho = 0.0 and 0.1, p-value∼0.39 and ∼10−05) and fruit fly (rho = −0.3 and−0.3, p-value∼10−14 and ∼10−14). In all cases we observe a nearly monotonic relationship, which is increasing for vertebrates, and decreasing for fruit fly.
Figure 9.
Number of exons is a component of isoform multiplicity co-occurrence (IMco), other species.
This figure is equivalent to Figure 7, but in this case we show the results for six other species (raw data with black dots, smoothed curve in blue and envelope in grey). We display the relationship of IMco vs. the product of human and mouse number of exons. We chose the product because it is a priori related to IMco through Equation (2) and Figure 5. From left to right and top to bottom we have the results for: chimpanzee (rho = 0.1, p-value = ∼10−02), cow (rho = 0.3, p-value = ∼10−17), rat (rho = 0.4, p-value = ∼10−32), chicken (rho = 0.3, p-value = ∼10−19), zebrafish (rho = 0.3, p-value = ∼10−14) and fruit fly (rho = 0.2, p-value = ∼10−10). Because the extent of isoform annotations goes from low to very low (only 5% of chimpanzee genes had more than one isoform), the relationships are weaker than in Figure 7.
Figure 10.
The interplay between isoform multiplicity co-occurrence (IMco) and protein divergence (PD) in the generation of human-mouse phenotypic differences.
Here we plot the same graph as in Figure 2A with two additions: a grey-shaded area corresponding to the 50%–70% zone that separates functional from non-functional PD (see Discussion); a red-shaded area below the black line, highlighting the fact that the latter is an upper threshold for IMco values obtained after exclusion of non-relevant isoforms. We can see that phenotypically-relevant IMco values (red-shaded area) are always lower than 1, indicating that differences in isoform multiplicity can contribute to human-mouse phenotypic differences all over the PD range. To the right of the grey-shaded area this contribution will be more relevant than that of PD; to the left both phenomena will very likely cooperate to generate phenotypic changes; the situation is unclear within the 50%–70% zone.