Ed's Big Plans

>>> Attached: ( parse_newick_xhtml.js | draw_phylogeny.js — implementation in JS | index.html — demo with IL-1 ) <<<
>>> Attached: ( auto_phylo_diagram.tgz — includes all above, a python script, a demo makefile and IL-1 data ) <<<

In this post, I discuss a Python script I wrote that will convert a Newick Traversal and a FASTA file into a browser-viewable interactive diagram using SVG+JS+XHTML. This method is compatible with WebKit (Safari, Chrome) and Mozilla (Firefox) renderers but not Opera. XHTML doesn’t work with IE, so we’ll have to wait for massive adoption of HTML5 before I can make this work for everyone — don’t worry, it’s coming.

To the left is what the rendered tree looks like (shown as a screen captured PNG).

I recommend looking at the demo index.xhtml above if your browser supports it now. The demo will do a far better job than text in explaining exactly what features are possible using this method.

Try clicking on the different nodes and hyperlinks in the demo — notice that the displayed alignment strips away sites (columns) that are 100% gaps so that each node displays only a profile relevant for itself.

I originally intended to explain all the details of the JavaScripts that render and drive the diagram, but I think it’d be more useful to first focus on the Python script and the data input and output from the script.

The Attached Python Script to_xhtml.py found in auto_phylo_diagram.tgz

I’ll explain how to invoke to_xhtml.py with an example below.

python to_xhtml.py IL1fasta.txt IL1tree2.txt "IL-1 group from NCBI CDD (15)" > index.xhtml

Arguments …

1. Plain Text FASTA file — IL1fasta.txt
2. Plain Text Newick Traversal file — IL1tree2.txt
3. Title of the generated XHTML — “IL-1 group from NCBI CDD (15)”

The data I used was generated with MUSCLE with default parameters. I specified the output FASTA with -out and the output second iteration tree with -tree2.

The Attached makefile found in auto_phylo_diagram.tgz

The makefile has two actual rules. If index.html does not exist, it will be regenerated with the above Python script and IL1fasta.txt and IL1tree2.txt. If either IL1fasta.txt or IL1tree2.txt or both do not exist, these files are regenerated using MUSCLE with default parameters on the included example data file ncbi.IL1.cl00094.15.fasta.txt. Running make on the included package shouldn’t do anything since all of the files are included.

The Example Data

The example data comes from the Interleukin-1 group (cd00100) from the NCBI Conserved Domains Database (CDD).

Finally, I’ll discuss briefly the nitty gritty of the JavaScripts found as stand-alones above that are also included in the package.

Part Two of Two

As is the nature of programatically drawing things, this particular task comes with its share of prerequisites.

In a previous — Part 1, I covered Interactive Diagrams Using Inline JS+SVG+XHTML — see that post for compatibility notes and the general method (remember, this technique is a holdover that will eventually be replaced with HTML5 and that the scripts in this post are not compatible with Opera anymore).

In this current post — Part 2, I’ll also assume you’ve seen Parsing a Newick Tree with recursive descent (you don’t need to know it inside and out, since we’ll only use these functions) — if you look at the example index.html attached, you’ll see that we actually parse the Newick Traversal using two functions in parse_newick_xhtml.js — tree_builder() then traverse_tree(). The former converts the traversal to an in-memory binary tree while the latter accepts the root node of said tree and produces an array. This array is read in draw_phylogeny.js by fdraw_tree_svg() which does the actual rendering along with attaching all of the dynamic behaviour.

Parameters ( What are we drawing? )

To keep things simple, let’s focus on three requirements: (1) the tree will me sketched top-down with the root on top and its leaves below; (2) the diagram must make good use of the window width and avoid making horizontal scroll bars (unless it’s absolutely necessary); (3) the drawing function must make the diagram compatible with the overall document object model (DOM) so that we can highlight nodes and know what nodes were clicked to show the appropriate data.

Because of the second condition, all nodes in a given depth of the tree will be drawn at the same time. A horizontal repulsion will be applied so that each level will appear to be centre-justified.

General Strategy

Because we are centre-justifying the tree, we will be drawing each level of the tree simultaneously to calculate the offsets we want for each node. We perform this iteratively for each level of depth of the tree in the function fdraw_tree_svg(). Everything that is written dynamically by JavaScript goes to a div with a unique id. All of the SVG goes to id=”tree_display” and all of the textual output including raster images, text and alignments goes to id=”thoughts”. To attach behaviour to hyperlinks, whether it’s within the SVG or using standard anchor tags, the “onclick” property is defined. In this case, we always call the same function “onclick=javascript:select_node()“. This function accepts a single parameter: the integer serial number that has been clicked.

Additional Behaviour — Images and Links to Databases

After I wrote the first drafts of the JavaScripts, I decided to add two more behaviours in the current version. First, if a particular node name has the letters “GI” in it, the script will attempt to hyperlink it with the proceeding serial number against NCBI. Second, if a particular node name has the letters “PDB” in it, the script will attempt to hyperlink the following identifier against RSCB Protein Database and also pull the first thumbnail with an img tag for the 3D protein structure.

Enjoy!

This was done because I wanted to show a custom alignment algorithm to my advisors. In this demo, I’ve stripped away the custom alignment leaving behind a more general yet pleasing visualizer. I hope that the Python script and JavaScripts are useful to you. Enjoy!

>>> Attached: ( inline_svg_demo.xhtml — demo using the below technique for a silly drawing ) <<<

Compatibility: The following method doesn’t work for Internet Explorer 8, so if interoperability is really important to you, skip this series! Don’t worry though — Firefox 4.0, Internet Explorer 9 and WebKit (Safari / Chrome) will all natively support HTML5. HTML5 includes inline SVG for HTML so this ad hoc version can be retired when these three products take hold. The below method is compatible with Firefox 3.x, Opera 9.x and the current version of WebKit (Safari 5.x, Chrome 7.x). For display reasons, I’ll show off bitmap files inline while displaying inline SVG in linked pages.

Part One of Two

Part one is for everyone! I describe how to display an SVG inline on a webpage (XHTML) along with some JavaScript driven behaviour (clicking changes a displayed message and CSS defined colours).

Part two is for the bioinformaticians and phylogenists that survive and stick around — I utilize a parsed Newick format traversal to build up the drawing of a tree recursively — then I finish off by reintroducing JavaScript driven interactivity — tree nodes are highlighted via CSS when clicked, and information is displayed for the selected node (perhaps a multiple sequence alignment, a global alignment score, branch lengths — etc.).

( See part two here: Simple Interactive Phylogenetic Tree Sketches JS+SVG+XHTML )

Wait a sec — what do you mean by inline SVG?

Scalable Vector Graphics (SVG) is an XML vector image format. All browsers allow you to specify SVG files to be included into a webpage as an image. What we’ll be discussing here is how to include an SVG image as an XML element directly in a XHTML file — that is, the single webpage file will contain all of the markup for itself and its included SVG images without linking to any external files.

I won’t go into detail about how to make SVG images. You may create them with software like InkScape or by manual manipulation of the raw XML using a text or code editor (which I’ve done in this post). If you do end up using drawing software, simply copy out all of the XML elements with a text editor afterward and inject them into the XHTML page you create.

XHTML File Nuances

Include the following as a attribute of your html tag — this ensures that the browser reading your page knows which document type definition (DTD) to use.

<html xmlns="http://www.w3.org/1999/xhtml">
...
</html>

You must include your SVG with the following tags.

<svg xmlns="http://www.w3.org/2000/svg" width="300" height="300"
    xmlns:xlink="http://www.w3.org/1999/xlink">
...
</svg>

In the above, we parameterize the “svg” tag with the correct xml namespace (xmlns) as an attribute. We specify the width and height of the image we want to inline and also specify that we want to be able to use hyperlinks within the SVG image with “xlink”. Notice that both namespace attributes are meant to enable a client browser to correctly render your page.

Smile Again!

A Silly Drawing and Target Behaviour

In this post, we’ll be manipulating a picture of a silly face I’ve named Smile Again — if you can render the attached file, the face looks like the one to the right.

The eventual target behaviour we want is as follows. Clicking on different parts of Smile Again’s face causes (1) the displayed message to change and (2) the clicked part to change colours. Notice that I’ll be using JavaScript (JS) to perform both of these functions. We’ll get into the specifics soon — if you’re feeling up to it, you can always go to the attached demo xhtml file now and look at the source to reverse engineer the thing.

We can create this face manually with the following inline SVG code — we’ll start with pure SVG and worry about the JavaScript behaviour afterward.

<svg xmlns="http://www.w3.org/2000/svg" width="300" height="300"
 xmlns:xlink="http://www.w3.org/1999/xlink">

<circle cx="100" cy="100" r="40"
 style="fill:white;stroke:black;stroke-width:2" />

<circle cx="100" cy="100" r="20"
 style="fill:black;stroke:black;stroke-width:2" />

<circle cx="200" cy="100" r="60"
 style="fill:white;stroke:black;stroke-width:2" />

<circle cx="200" cy="100" r="10"
 style="fill:black;stroke:black;stroke-width:2" />

<path d="M60 200 C80 300 220 300 240 200 L60 200"
 style="fill:white;stroke:black;stroke-width:2" />

<path d="M100 235 L200 235"
 style="fill:white;stroke:black;stroke-width:2" />

<path d="M130 224 L130 246"
 style="fill:white;stroke:black;stroke-width:2" />

<path d="M150 220 L150 250"
 style="fill:white;stroke:black;stroke-width:2" />

<path d="M170 224 L170 246"
 style="fill:white;stroke:black;stroke-width:2" />

</svg>

Each of the XML elements above correspond to a different piece of the SVG image. The first four circles draw the eyes, next the complicated path draws the mouth and the four simpler paths draw the stitches of the teeth.

Changing the Displayed Message with JavaScript

Now that we’ve drawn Smile Again, we want to have it react to mouse clicks. I want to have both the clicked features highlight as well as the displayed message to change. This would be an important feature of actual practical diagrams. To make this task a bit easier to explain, I will break down the behaviour into smaller components and functions and build them back up as I go along. This will also make it a lot easier to read.

Let us focus on just the two circles that makes up Smile Again’s right eye (on the left side of the diagram).

We want our clicks to call a function — in order for that to happen, we need to use anchor tags (hyperlinks) within the SVG. This is made possible with our careful xmlns (namespace) attributes we declared earlier!

<svg xmlns="http://www.w3.org/2000/svg" width="300" height="300"
    xmlns:xlink="http://www.w3.org/1999/xlink">

<a xlink:href="#" onclick="javascript:change_thoughts('Oww! My right cornea!')">
    <circle cx="100" cy="100" r="40"
        style="fill:white;stroke:black;stroke-width:2" />
</a>

<a xlink:href="#" onclick="javascript:change_thoughts('Ouch! My right iris!')">
    <circle cx="100" cy="100" r="20"
        style="fill:black;stroke:black;stroke-width:2" />
</a>
...
</svg>

Notice that we’ve now enclosed entire clickable elements in anchor tags and have used the “xlink:href” hyperlink attribute. In reality, we can have these links point to actual other webpages. Instead, these point to the same page and call the JavaScript function to change the text in the description called “change_thoughts”. In this case, the output is an expression of pain after being poked in the eye (“Oww! My right cornea!”). We can now define the “change_thoughts” function and also the place where we want our messages to appear.

<head>
...
function change_thoughts(say_this) {
    document.getElementById("thoughts").innerHTML = "\"" + say_this + "\""
}
...
</head>
<body>
...
<h2 id = "thoughts" style="font-family:monospace;">
</h2>
...
</body>

The “change_thoughts” function can go in the head or somewhere early in the body of the XHTML. It doesn’t really matter as long as it occurs before the SVG. This function gets an element with the id “thoughts” and changes the HTML inside of it. In the above, I add a pair of double quotes to the string to show. The element with the id “thoughts” is a h2 element — this really could been any valid HTML element. You do however have to place the “thoughts” element somewhere after  you declare the “change_thoughts” function.

Changing the Highlighting with JavaScript and CSS

The last thing we want to do is to change the highlighting by making our JavaScript function touch the style (CSS) of the inline SVG element that was clicked. We could use a separate CSS file, but in the spirit of making everything one inlined file, the CSS is specified with the “style” html attribute.

Again, we’ll be looking only at the two circles corresponding to Smile Again’s right eye (I’ve changed the way the lines are broken in the code listing below so that it will actually fit on this page).

<a xlink:href="#"
    onclick=
"javascript:change_thoughts('Oww! My right cornea!','right_cornea', '#F00')">
    <circle cx="100" cy="100" r="40"
        style="fill:white;stroke:black;stroke-width:2"
        id="right_cornea" />
</a>

<a xlink:href="#"
    onclick=
"javascript:change_thoughts('Ouch! My right iris!', 'right_iris', '#00F')">
    <circle cx="100" cy="100" r="20"
        style="fill:black;stroke:black;stroke-width:2"
        id="right_iris" />
</a>

Here are our two changes. First — each of the SVG elements now have a unique id (“right_cornea”, “right_iris”) so that we can use JS to manipulate each shape independently. Second — we’re now calling a new version of “change_thoughts” — this time, we’re giving the function the id of the shape to highlight and the colour (in hexadecimal) to highlight it with.

Finally, I can show you the new version of “change_thoughts” which will accept the new arguments and change the colours of the shapes.

function change_thoughts(say_this, highlight_this, colour) {

    // Say this ...
    document.getElementById("thoughts").innerHTML = "\"" + say_this + "\""

    // Highlight this ...
    var change = document.getElementById(highlight_this)
    change.setAttribute("style",
        "fill:" + colour + ";stroke:black;stroke-width:5")
}

In the above, we’ve added the code after “// Highlight this …”. We select the element “highlight_this” — remember, the id and colour are given by the function called with mouse clicks in the SVG. We then change the style by overwriting the inline CSS with the new specified fill colour (and a fat black stroke).

We now have a problem — each new mouseclick on a different shape will cause that new shape to highlight. The previous highlighted shape doesn’t automatically revert, and eventually all of the shapes are highlighted. What we need to do now is to improve the “change_thoughts” function one last time with some way to “undo” the previous change, so that new calls to this function will not only highlight the new shape, but will also remove the highlight from the previous shape.

Finishing with an Undo Object to remove previous changes

In the following version, we add an object to behave like a hash in the JavaScript. This hash will take html id attributes as keys, and the previous style as values. This behaves like an “undo” stack since we apply the styles to the elements with the corresponding id before we commit the changes to the new shape.

var _undo_render_hash = new Object

function change_thoughts(say_this, highlight_this, colour) {

    // Say this ...
    document.getElementById("thoughts").innerHTML = "\"" + say_this + "\""

    // Undo previous highlight ...
    for(var i in _undo_render_hash) {
        document.getElementById(i).setAttribute("style", _undo_render_hash[i])
    }

    // Reset undo stack ...
    _undo_render_hash = new Object

    // Highlight this ...
    var change = document.getElementById(highlight_this)
    _undo_render_hash[highlight_this] = change.getAttribute("style")
    change.setAttribute("style",
        "fill:" + colour + ";stroke:black;stroke-width:5")
}

Walking through the code, we start with a blank object as the undo stack — this is OK. On the first call of this function, since the undo object is empty, simply nothing is done before the new shape is highlighted. After the first call, all of the changes that were committed in the last call of the function are undone as their original styles are reapplied. We just add the next object to change to the undo object before applying changes to it.

The undo object is treated like a hash here, and in fact, we iterate all of its elements even though we only expect to have captured one id in the last call of the function. This means we can use this undo technique for functions that will modify many shapes with many changes since each of those modifications is iteratively undone in the next call. We must of course ensure that we save the original style of each of these shapes.

Next

At some point, I’ll want to revisit this and combine some of the other JavaScript stuff I’ve shown in this blog before. Displaying a rendered clickable phylogenetic tree is a task I figured out since I needed to quickly visualize some data for my team in my thesis work. Since I’ve already covered parsing a newick tree in this blog, it makes sense to complete the discussion with how I ended up with my phylogeny visualizer.