phase_subgenomes.md

Subgenome-aware analyses

Let’s delve into a detailed explanation of how we can phase subgenomes corresponding to the balsa tree, cotton, heritiera littoralis, and durian using the Chr3 of cacao as an example.

Phased subgenomes for the balsa tree.

We start by using the WGDI toolkit (Sun et al. 2022) with the parameter '-d' to plot a homologous gene dot plot between the balsa tree and cacao. The parameter ‘ancestor_top = tc180s.ancestor.txt’. Only the Chr3 of cacao is colored; all other chromosomes are white. The resulting dot plot is shown in the figure below.

tc180s.ancestor.txt

1	1	3130	white	1
2	1	2675	white	1
3	1	2367	#ff9900	1
4	1	2285	white	1
5	1	2366	white	1
6	1	1751	white	1
7	1	1253	white	1
8	1	1496	white	1
9	1	2718	white	1
10	1	1224	white	1

Next, we use WGDI with the '-icl' and '-bi' parameters to find and integrate collinear genes between them. This allows us to easily screen for collinear fragments. We then set parameters (multiple = 1, homo = 0.5, 1) to extract synteny blocks, which are mostly composed of red dots, from the comparison of the balsa tree and cacao using WGDI with the '-c' parameter. We then use WGDI with the '-bk' parameter to examine whether the filtered synteny blocks are reasonable. Since we haven’t calculated the Ks values, they are all defaulted to -1. This step is absolutely crucial because it ensures that the balsa tree's chromosomes correspond to the unique homologous region of cacao. The filtered results are shown in the figure below.

Finally, we use the '-km' parameter of WGDI to extract the mapping of cacao colors to the balsa tree, resulting in the ancestor_file of the balsa tree, namely opy300s.ancestor.txt. We then generate a dot plot using the '-d' parameter in WGDI, specifying ‘ancestor_left = opy300s.ancestor.txt’. The resulting dot plot is shown in the figure below.

Omitting the chromosomes marked as white, the remaining chromosomes are as follows.

opy300s.ancestor.txt

1	1	1375	#ff9900	1
15	1	807	#ff9900	1
26	1	829	#ff9900	1
5	1	1655	#ff9900	1
6	1	1585	#ff9900	1

We assign all balsa chromosome blocks to one of five subgenomes based on chromosome complementarity. These five chromosomes independently correspond to Chr3 of cacao, indicating that they have not undergone any fusion or fission events. As a result, they can be directly phased into five distinct subgenomes. The last column of ancestor_file represents different subgenomes.

opy300s.ancestor.new.txt

1	1	1375	#ff9900	1
15	1	807	#ff9900	2
26	1	829	#ff9900	3
5	1	1655	#ff9900	4
6	1	1585	#ff9900	5

opy300s.ancestor.txt

We record the subgenome regions on the chromosomes and use WGDI with the parameters '-pc' and '-a' to obtain hierarchical gene lists. The resulting dot plot is shown in the figure below.

The hierarchical gene lists : tc180s_opy300s.alignment.csv

tc180s3g00001
tc180s3g00002
tc180s3g00003
tc180s3g00004
tc180s3g00005			opy300s6g00577		opy300s26g00710
tc180s3g00006	opy300s1g00373		opy300s6g00576	opy300s15g00208	opy300s26g00709
tc180s3g00007	opy300s1g00374		opy300s6g00574	opy300s15g00207	.
tc180s3g00008	opy300s1g00375		opy300s6g00568	.	.
tc180s3g00009	opy300s1g00376	opy300s5g01075	opy300s6g00567	opy300s15g00206	opy300s26g00708
tc180s3g00010	opy300s1g00377	opy300s5g01074	opy300s6g00393	.	opy300s26g00707
tc180s3g00011	opy300s1g00379	opy300s5g01073	.	.	.
tc180s3g00012	.	opy300s5g01072	.	opy300s15g00205	.
...	...	...	...	...	...

Phased subgenomes for Heritiera littoralis.

tc180s_hli2s.alignment.csv

tc180s3g00001
tc180s3g00002
tc180s3g00003
tc180s3g00004
tc180s3g00005		hli2s13g00417
tc180s3g00006	hli2s12g00507	hli2s13g00418
tc180s3g00007	hli2s12g00508	hli2s13g00420
tc180s3g00008	.	.
tc180s3g00009	hli2s12g00510	.
tc180s3g00010	hli2s12g00511	.
tc180s3g00011	.	.
tc180s3g00012	hli2s12g00512	hli2s13g00422
...	...	...

Phased subgenomes for durian.

tc180s_dz186s.alignment.csv

tc180s3g00001
tc180s3g00002
tc180s3g00003
tc180s3g00004
tc180s3g00005	dz186s3g01240
tc180s3g00006	dz186s3g01239	dz186s14g00867
tc180s3g00007	dz186s3g01238	.	dz186s15g00824
tc180s3g00008	.	dz186s14g00866	.
tc180s3g00009	dz186s3g01236	dz186s14g00865	dz186s15g00823
tc180s3g00010	dz186s3g01235	dz186s14g00864	.
tc180s3g00011	dz186s3g01234	.	.
tc180s3g00012	dz186s3g01233	dz186s14g00863	dz186s15g00822
...	...	...	...

Relationships among the subfamilies

Let’s discuss how we can infer relationships among the subfamilies from collinear genes.

First, we combine the subgenomes of different species into a single file. Since cacao serves as the reference for all of them, we can directly paste them into a file. Column 1 is from cacao, Columns 2 and 3 are from heritiera littoralis, Columns 4, 5, and 6 are from durian, and Columns 7, 8, 9, 10, and 11 are from the balsa tree.

tc180s3g00001
tc180s3g00002
tc180s3g00003
tc180s3g00004
tc180s3g00005		hli2s13g00417	dz186s3g01240					opy300s6g00577		opy300s26g00710
tc180s3g00006	hli2s12g00507	hli2s13g00418	dz186s3g01239	dz186s14g00867		opy300s1g00373		opy300s6g00576	opy300s15g00208	opy300s26g00709
tc180s3g00007	hli2s12g00508	hli2s13g00420	dz186s3g01238	.	dz186s15g00824	opy300s1g00374		opy300s6g00574	opy300s15g00207	.
tc180s3g00008	.	.	.	dz186s14g00866	.	opy300s1g00375		opy300s6g00568	.	.
tc180s3g00009	hli2s12g00510	.	dz186s3g01236	dz186s14g00865	dz186s15g00823	opy300s1g00376	opy300s5g01075	opy300s6g00567	opy300s15g00206	opy300s26g00708
tc180s3g00010	hli2s12g00511	.	dz186s3g01235	dz186s14g00864	.	opy300s1g00377	opy300s5g01074	opy300s6g00393	.	opy300s26g00707
tc180s3g00011	.	.	dz186s3g01234	.	.	opy300s1g00379	opy300s5g01073	.	.	.
tc180s3g00012	hli2s12g00512	hli2s13g00422	dz186s3g01233	dz186s14g00863	dz186s15g00822	.	opy300s5g01072	.	opy300s15g00205	.
tc180s3g00013	hli2s12g00513	.	.	.	dz186s15g00821	.	opy300s5g01071	.	.	.

Next, using the hierarchical gene lists, we infer maximum likelihood (ML) trees using IQ-TREE through WGDI with the parameter '-at'. We then use ASTRAL-III v.5.7.8 with the parameter ‘-t 16’ to construct the coalescent tree and estimate branch support.

The collinear gene trees are tc_hli_dz_opy.nwk and the coalescent tree is tc_hli_dz_opy.tre

We can rearrange the initial arbitrary arrangement of phase subgenomes that differentiates species. By adjusting the order of phase subgenomes within the same species, we find that phase subgenomes S1 and S2 for Heritiera littoralis, durian, and balsa tree form separate clusters. Similarly, phase subgenomes S1, S2, and S3 for durian and balsa tree also form separate clusters.

Finally, we adjust the order of the subgenomes in the ancestor_file. For instance, the phased subgenomes for the balsa tree are updated accordingly.

1	1	1375	#ff9900	3
15	1	807	#ff9900	4
26	1	829	#ff9900	5
5	1	1655	#ff9900	1
6	1	1585	#ff9900	2

We continue to use WGDI with the parameters '-pc' and '-a' to obtain new hierarchical gene lists.

Phased subgenomes for cotton.

We used the WGDI toolkit with the parameter '-d' to plot homologous gene dot plot between balsa tree and cacao. The mapping of cacao's Chr3 to cotton chromosomes is dispersed, making it difficult to partition directly as before. In this case, we can use complementary combinations between synteny blocks to determine the phase of different subgenomes. Specifically, blocks from cotton Chrs 12 and 13 correspond intact to cacao's Chr3 and can be considered separate subgenomes, each marked with a different color. The blocks from Cotton Chrs 7 and 8 complement Chr3 of cacao and form another subgenome. cotton Chrs 1 and 3 are complementary, and the upper region of cotton Chr9 aligns perfectly with the remaining blocks, making them another subgenome. The remaining blocks of cotton Chr9 constitute yet another subgenome. This way, we phased these blocks into five subgenomes and grouped them randomly.

1	1691	2175	#ff9900	1
1	2530	2841	#ff9900	1
3	1256	1292	#ff9900	1
4	2106	2145	#ff9900	1
9	518	686	#ff9900	1
7	475	828	#ff9900	2
8	1112	1370	#ff9900	2
8	1740	2165	#ff9900	2
9	1846	1936	#ff9900	3
9	2008	2236	#ff9900	3
9	2763	2840	#ff9900	3
11	3659	4131	#ff9900	4
12	1	1101	#ff9900	5

We continue to use WGDI with the parameters '-pc' and '-a' to obtain hierarchical gene lists.

As we've previously categorized the balsa tree into five subgenomes, we combined these subgenomes with those of cotton into a single file. Column 1 is from cacao, Columns 2, 3, 4, 5, and 6 are from the balsa tree, Columns 7, 8, 9, 10, and 11 are from cotton. We then used WGDI with the '-at' parameter and ASTRAL-III version 5.7.8 to infer relationships among the subfamilies.

By rearranging the order of phased subgenomes in cotton, we observed a one-to-one relationship between the five subgenomes of the balsa tree and cotton.

1	1691	2175	#ff9900	3
1	2530	2841	#ff9900	3
3	1256	1292	#ff9900	3
4	2106	2145	#ff9900	3
9	518	686	#ff9900	3
7	475	828	#ff9900	4
8	1112	1370	#ff9900	4
8	1740	2165	#ff9900	4
9	1846	1936	#ff9900	1
9	2008	2236	#ff9900	1
9	2763	2840	#ff9900	1
11	3659	4131	#ff9900	2
12	1	1101	#ff9900	5

Additionally, we can quickly phase subgenomes based on the sequence similarity between the balsa tree and cotton. Because red dots represent high sequence similarity, synteny blocks with more red dots are assigned to their corresponding subgenomes. Importantly, both methods yielded consistent results.

Furthermore, subgenomes could also be phased based on additional information, such as Ks values, structural consistency, and the number of retained genes. Regardless of the method employed for subgenome phasing, we ultimately validated the results using colinear gene trees.

We merged all the corrected hierarchical gene lists and subsequently reconstructed a consensus tree using WGDI with the '-at' parameter and ASTRAL-III version 5.7.8.