mt-Tree Mutation Timeline

[PeterLarsen]

Why do you want to destroy the actual mutations from my tree with just garbage/random numbers to create an alternative X-axis? Do you think my X-axis is wrong? My X-Axis is based on number of mutations from mt-Eve directly taken from mt-Tree. It's unit it in mutations.

For chopping off, yes it does chops off and people still live today under that clade - because, it exists as private mutations and there will not be enough population to create/declare a new branch. This does not mean new branches don't exist. It means, there are no people living today having only those defining mutations for that parent clade.

If you feel you can extend it and provide an alternative versions, feel free to post it in another thread, not this - as I want to stick with OP.

My point was to offer a way to look at your version in conjunction with an alternative view of the tree. That will enable you to not being visually blinded by faulty conclusions.

What in your version of the tree looks like this:

        66│                                                     └[?]2b1
        68│                                                            └[?]2b1a
        69│                                                              ├[?]2b1a2	 
        69│                                                              └[?]2b1a3

Could in reality, with a correct X-axis timeline, look more like this (an X] represents a subclade that dies):

───────┬─────────────────────────┬────────────────┬──────────────────────┬────────────────────┬───────────────┬──────────────────┬─────────────────┬────── [?]2b1a
       │                         │                └[?]2b1a[?]──X]        │                    └[?]2b1a[?]──X] │                  └[?]2b1a[?]──X]   │
       │                         │                                       └[?]2b1a[?]──X]                      └────[?]2b1a[?]──X]                  └────── [?]2b1a2
       │                         │
       │                         └──────────────────────────────────────────────── [?]2b1a[?]───────X]
       └────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────── [?]2b1a3

X-axis could represent 10000 years, for example. Close (in time) to the spawn from the parent, they are all (almost) GD 1 from parent, but they may have been created with more than 1000 years between them. Subclades with a GD of 1 from parent does not automatically mean they are spawed in the same century


Using your visualization you don't know when, in time, the subclades spawed from the parent. You don't know how many have spawed, and you dont know how many and when spawned subclades died out (subclades can die out 1000 of years later).
You don't know how big [?]2b1a is (in amount of parallell maternal lines) during the time span, the size can vary up and down => spawning different amount of subclades / time span at different historical times.

What you are trying to do corresponds to figuring out x from an equation something like this:
(x + 5y)/z = (3p * q/7) ^ r

... where y, z, p, q and r is unknown. In other words: you will find what you are looking for, no matter what you are looking for. The amount of possibilities are endless. Please stop trying to draw conclusions from such
thin data. I offered a way to look at it using more data, I am sorry to hear that you think I "want to destroy the actual mutations from my tree with just garbage/random numbers". That was not my intention. My intension was
to stress the fact that the X-axis in your tree does not represent a scaled timeline.

One common mistake when looking at simplified trees is to look at mutations / SNPs as breakpoints dividing a "line" into two separate lines. One line in the simplified version can actually be hundred or thousand parallel lines. It helps visualizing a mutation as occurring "on the lines" rather than at the breakpoints.

[felix]

My point was to offer a way to look at your version in conjunction with an alternative view of the tree. That will enable you to not being visually blinded by faulty conclusions.

Thank you for the offer. If you have an alternate better version of the tree to contribute by extending what I did, please start a new thread.

[MMaddi]

Thank you for the offer. If you have an alternate better version of the tree to contribute by extending what I did, please start a new thread.

I understand his point, which goes to the actual way that mutations occur and create new subclades, and it sounds valid to me. The fact that his point undercuts your method and its conclusions does not make it off topic.

The essence of the scientific method is to come up with a theory and then submit it for other scientists or interested layman to consider for possible corrections or improvement. It's not the scientific method to say that differing ideas about your theory are not allowed in your thread.

[PeterLarsen]

Thank you for the offer. If you have an alternate better version of the tree to contribute by extending what I did, please start a new thread.

Inspired by the challenge to create a better view of a 2D MtDNA-tree where y-axis is Haplogroup/subclade and x-axis is the amount of currently known defining mutations, here comes a suggestion:



It is supposed to be easier to understand that the actual mutations does not occur where the lines divide, but somewhere "on the lines". Limited by a screen character width it seems some gaps seems very small, but that is the result when trying to shrink tens of thousands years into less than 100 character in width presentation .

I guess it comes down to personal opinion what is the best visualization for this sort of view. Personally I prefer not getting tricked by tree-structures that does not summarize a realistic ensemble regarding the time line. Since it is impossible to perfectly date when a subclade spawnes from its parent clade (when talking about the bigger clades, excluding recent young clades) I prefer not trying to define what is not possible to define. My visualization is not perfect, still parallel subclades appears to be spawn in the same time frame, which is of cource not realistic.

Compared to your visualization, mine don't present the illusion of parent clades getting extinct in the moment subclades is created, and the blocks that is supposed to be equal to mutations is better visualized and not hidden by the clade name. One mutation is represented by a division of the simplified line, while 2 or more mutations is marked out by the black blocks.

One thing to possibly improve is to draw separate lines from parent clades to the subclades all over the parent clade x-axis, like my example in my previous post, and that way better visualizing that parallel subclades from the same parent clade can be separated by an enormous long time span. The drawback from such a visualization is that the screen will be filled with lines and difficult to overview. It is important though to keep in mind that single lines in the simplificated tree can represent everything from a single maternal line to thousands of parallel lines.

Looking at the tree from the perspective of present time, one must also keep in mind that there have been an enourmous amount of "private" subclades spawned from all parts of the tree in all historical time that hasn't survived and did not create a new subclade that did survive until present time. Only a fraction of all mutations survive due to the mechanics of evolution.

I have left all clade names far to the right, representing present time, for the reason that you can't know for sure when in time the subclade was spawn only by looking at the amount of mutations the clade has. One other thing to note is that the X-axis represent a timescale BUT pay attention it is not to scale. Just because one subclade is more to the right than another one, that doesn't mean the clade must be younger. The timescale is only normalized and in order withing the same line between MtEVE and one specific haplogroup clade in present time. One "character" in the visualization can represent 1 generation in one line while another character at the same X position on a different line can represent 100 generations, representing a totally different historical time. In short that means that locally groups or signature patterns can't be assumed to have happened in a short period of time. If, for example, one clade is defined by 10 mutations and then spawns into 10 different parallel subclades, you can't from that limited data know if that happened in a 200 year time frame or a 2000 year time frame. Depending on the size of the parent clade for the actual time span, in combination with other unknown factors, there could be very different causes for this result, and without combining the data with other data, it is impossible to draw any conclusion. If, however, one add information and predict when the actual subclades is spawn, one could normalize the X-axis with each other, and that way other possibilities in drawing conclusion from signature patterns can be drawn. That requires a tremendous work though, and the precision will vary from subclade to subclade.

[PeterLarsen]

Inspired by the challenge to create a better view of a 2D MtDNA-tree where y-axis is Haplogroup/subclade and x-axis is the amount of currently known defining mutations, here comes a suggestion:



...

One thing I forgot to mention is that it would be possible to better visualize it by reordering the parallel subclades so that the subclades with higher GD to root (of all children subclades) spawns further to the left than subclades with lower GD. I refer to parallel subclades that shares the same parent, those who are in line vertically in the current visualizations. That would better normalize the X-axis, but still not a reliable scaled time, of course.

If I figure out a simple way to programmatically create such a view, I will show what I mean. It may introduce some difficulties into the logic of my tool though, so I might not be able to do it.

[felix]

Please stop trying to draw conclusions from such
thin data. I offered a way to look at it using more data, I am sorry to hear that you think I "want to destroy the actual mutations from my tree with just garbage/random numbers". That was not my intention. My intension was to stress the fact that the X-axis in your tree does not represent a scaled timeline.

Based on what you had posted, you destroyed the entire purpose of what I posted by reducing the actual mutation scaled timeline from build-16 mt-tree with garbage/random numbers.

Just to myself: I keep forgetting Pro 26:4 esp., with posting on online forums. No offence intended.

[N21163]

Just to myself: I keep forgetting Pro 26:4 esp., with posting on online forums. No offence intended.

Curious, if this reference is meant to be "just for yourself" why post it?

[felix]

Curious, if this reference is meant to be "just for yourself" why post it?

It is intended for curious people like you

[MMaddi]

Based on what you had posted, you destroyed the entire purpose of what I posted by reducing the actual mutation scaled timeline from build-16 mt-tree with garbage/random numbers.

Just to myself: I keep forgetting Pro 26:4 esp., with posting on online forums. No offence intended.

This reminds me of George Costanza - http://www.youtube.com/watch?v=1TsdXa_7kzs

[felix]

This reminds me of George Costanza - http://www.youtube.com/watch?v=1TsdXa_7kzs

Thanks for you unwanted and useless response.

[MMaddi]

Thanks for you unwanted and useless response.

You're quite welcome!

[N21163]

It is intended for curious people like you

Doubtful

[N21163]

You're quite welcome!

Pertinent Seinfeld reference

[PeterLarsen]

I see now that I forgot to link to the modified version:

http://artificial.se/DNA/mtdnatree.txt

And a thought that crossed my mind: wouldn't it better to actually create the tree from right to left and have all clades perfectly lined up vertically rightmost supposed to represent present? Then we move back in time and when the mutations in one subclade "fall short" align the subclude to the parent at that position. That way we move from more certain (rightmost) back in time on to the more hypothetical mtEVE (leftmost) ?

Well, we will find out fast that the 1 mutation per X years formula fall short since we will not end up with a proper structured tree, but that is of course obvious.

What I mean is instead of this:

        21│ │        └♦♦♦♦♦♦♦♦♦♦┬──────────────────────────────────────────────────────────────── L0d                   [  11]
        25│ │                   ├♦♦♦┬──────────────────────────────────────────────────────────── L0d1'2                [   4]
        30│ │                   │   ├♦♦♦♦┬─────────────────────────────────────────────────────── L0d1                  [   5]
        31│ │                   │   │    ├┬────────────────────────────────────────────────────── NL:L0d1               [   1]
        38│ │                   │   │    │├♦♦♦♦♦♦┬─────────────────────────────────────────────── L0d1a                 [   7]
        41│ │                   │   │    ││      └♦♦───────────────────────────────────────────── L0d1a1                [   3]
        39│ │                   │   │    │└♦♦♦♦♦♦♦┬────────────────────────────────────────────── L0d1c                 [   8]
        45│ │                   │   │    │        └♦♦♦♦♦┬──────────────────────────────────────── L0d1c1                [   6]
        46│ │                   │   │    │              ├──────────────────────────────────────── L0d1c1a               [   1]
        46│ │                   │   │    │              └──────────────────────────────────────── L0d1c1b               [   1]
        36│ │                   │   │    └♦♦♦♦♦┬───────────────────────────────────────────────── L0d1b                 [   6]
        40│ │                   │   │          └♦♦♦────────────────────────────────────────────── L0d1b1                [   4]
        31│ │                   │   └♦♦♦♦♦┬────────────────────────────────────────────────────── L0d2                  [   6]
        32│ │                   │         ├┬───────────────────────────────────────────────────── L0d2a'b               [   1]
        42│ │                   │         │├♦♦♦♦♦♦♦♦♦┬─────────────────────────────────────────── L0d2a                 [  10]
        43│ │                   │         ││         └─────────────────────────────────────────── L0d2a1                [   1]
        46│ │                   │         │└♦♦♦♦♦♦♦♦♦♦♦♦♦──────────────────────────────────────── L0d2b                 [  14]
        45│ │                   │         └♦♦♦♦♦♦♦♦♦♦♦♦♦───────────────────────────────────────── L0d2c                 [  14]
        39│ │                   └♦♦♦♦♦♦♦♦♦♦♦♦♦♦♦♦♦─────────────────────────────────────────────── L0d3                  [  18]

We can visualize it like this:

        21│ │                                   ────────────♦♦♦♦♦♦♦♦♦♦┬───────┬────────────────── L0d                   [  11]
        39│ │                                                         │       └♦♦♦♦♦♦♦♦♦♦♦♦♦♦♦♦♦─ L0d3                  [  18]
        25│ │                                                         └♦♦♦┬────────────────────── L0d1'2                [   4]
        30│ │                                                             └┬♦♦♦♦┬─────┬────────── L0d1                  [   5]
        36│ │                                                              │    │     └♦♦♦♦♦┬──── L0d1b                 [   6]
        40│ │                                                              │    │           └♦♦♦─ L0d1b1                [   4]
        31│ │                                                              │    └┬────┬────────── NL:L0d1               [   1]
        38│ │                                                              │     │    └♦♦♦♦♦♦┬─── L0d1a                 [   7]
        41│ │                                                              │     │           └♦♦─ L0d1a1                [   3]
        39│ │                                                              │     └♦♦♦♦♦♦♦┬─────── L0d1c                 [   8]
        45│ │                                                              │             └♦♦♦♦♦┬─ L0d1c1                [   6]
        46│ │                                                              │                   ├─ L0d1c1a               [   1]
        46│ │                                                              │                   └─ L0d1c1b               [   1]
        31│ │                                                              └♦♦♦♦♦┬─────────────── L0d2                  [   6]
        32│ │                                                                    ├┬────────────── L0d2a'b               [   1]
        42│ │                                                                    │├───♦♦♦♦♦♦♦♦♦┬─ L0d2a                 [  10]
        43│ │                                                                    ││            └─ L0d2a1                [   1]
        46│ │                                                                    │└♦♦♦♦♦♦♦♦♦♦♦♦♦─ L0d2b                 [  14]
        45│ │                                                                    └─♦♦♦♦♦♦♦♦♦♦♦♦♦─ L0d2c                 [  14]

I know, the model suffers from approximately the same issues, we can't predict when a subclade is branched from its parent only by looking at the mutation counts, but at least this visualization have the present time vertically aligned.

After a model like the above suggestion is created for the whole tree one can normalize the X-axis and the result should represent a better approximation, I believe. Still, we have lost the exact 1 char = 1 mutation = x number of years, but that relation is lost as soon as we move backward in time from the present and tries to combine the subclades into its parents.

And yes, don't worry. There are no random numbers introduced when using this method to create a tree from mutation counts.

What are your comment on such a visualization Felix, wouldn't it be a good alternative representation to get a nuanced perception of your concept?

[PeterLarsen]

Giving it some minutes more thought, we actually don't need extra rows for the unnamed branches. As I have moved away from the fixed, and yet unrealistic, 1 char per mutation "scale" I can skip using the breakpoint between parent and subclades as a mutation mark.

That leave us with the following example:



The former no labels and summon labels: L0d1'2, NL:L0d1, L0d2a'b are now represented with dots for the actual mutations only. It saves us a lot of lines.