Thread Closed 
Week 2: Slicing the wheat genome
09-04-2014, 09:15 PM
Post: #1
Week 2: Slicing the wheat genome
Taken from Science Magazine Article: Slicing the wheat genome Slicing the wheat genome

This visualization is network graph showing the genome dominance between different types of wheat cells at different stages of maturation. Specifically, this network represents cell types grouped into categories representing cell function with color to show data. For such a large visualization, a surprisingly small amount of information can be realized given what seems to be a large amount of encoded data.

The visualization is shown twice to contrast cell type and development (left) and genome dominance (right). The inset shows the network with major connective hubs outlined in red. The items, or nodes in the visualizations represent a cell where each cell is categorized function. For the left visualization, the cell development attribute is encoded into multiple different colors. For the right visualization, cell genome dominance is encoded by green, yellow, and purple color gradient.

The use of color in this visualization is important because that is how the viewer is able to discern cell dominance and developmental stages. Given the large number of nodes and links, there is little space left to show any other information about each cell. For the left visualization, the solid, well defined colors effectively portray the intended information, though, the light pink and dark pink are too similar. However, the gradient used in the right visualization is horrible. With such a large number of nodes, the colors often run together, and nodes get covered up. For such a large and busy visualization, only high level conclusions about the data can be achieved.

Another criticism of this visualization is the unnecessary size of the category labels. The labels overlap the visualization with a black line bordered by a solid white border. Not only does this distract from the visualization, but the lines cover up the data. The authors could have easily removed the useless white borders and resized the labels. Similarly, the inset also distracts from the main visualization. This inset could be removed and the information incorporated into the main network.

This visualization type does seem ill suited in a static medium. The author is attempting to encode too much data in a small space, using only colors and categories. A network of this size would be better suited in an interactive environment where the viewer can mouse over nodes to view data, zoom in on categories, or even highlight link paths. Overall, it would seem that using a maximal amount of ink was the focus of the design, and effectively portraying the data was just an afterthought.

Attached File(s) Image(s)
Find all posts by this user
09-05-2014, 04:01 PM
Post: #2
RE: Week 2: Slicing the wheat genome
I certainly agree that the only high level conclusions can be drawn from the data. The graphic is very cluttered but if the authors are just trying to establish broad connections between different gene ontologies I think it serves this purpose well. The links between major groups are more or less visible, and each group is clearly labeled by the large middle ontology labels. I don't mind the large space that they take up though because only high level analysis can be done from this graph. Loosing a little bit of resolution for label clarity is a fair trade off in my opinion. I think the visualization could be improved upon by using enclosure to further separate out the categories. They are labeled well but if the author were to circle regions in a light color it might help the groups and connections between them pop in my mind. What do you think?
Find all posts by this user
09-07-2014, 04:38 AM
Post: #3
RE: Week 2: Slicing the wheat genome
For a person with limited knowledge of biology, on first glance it is very difficult to comprehend the intended representation. After spending some time analysing the visualization and reading through the linked paper I was able to understand what the main focus was about. The visualization is indeed quite great as 2 figures are conveying a lot of information with such limited text. An individual can definitely make a rough estimate about the different developmental stages as well as the dominance percentage of the wheat cells by reading the centre text and legends.

Few points which I thought could be used towards improving the visualization:

Point 1: (Legend): The triangular legend on the right is slightly confusing. I do agree looking at the colors on can determine approximate dominance percentage of a particular type of genome. However lack of numerical data prevents us from deducing any concrete statistics while linking the colors with their corresponding dominance percentages.

Point 2: (Pointers) I completely agree with (theBlur05) and the bold pointers seem to be quite distracting. I also felt that the pointers should not overlap the main content of the visualisation (cells in our case). Perhaps a better alternative could be adding the pointers outside on the white background and just link them to the specific areas at the ends instead of overlapping the cells.

Point 3: (Extra figure): The third figure does not add much significant data to the visualization and I felt as compared to ratio of space occupied by it vs amount of information conveyed , it seems to be an unnecessary addition.

Point 4: (Color Scheme): The portions of the visualization where a lot of cells overlaps. I thought it would better to use a different color scheme rather than using similar shades of a one color. Also I do agree with (nbertagnolli) and feel that it would be apt to indicate the overlapping cells in the better way which could help us easily distinguish different cells dominance percentages .
Find all posts by this user
09-07-2014, 02:06 PM
Post: #4
RE: Week 2: Slicing the wheat genome
If the baseline of the image is to show some abstract information of cells and genome, I think the purposes is well fulfilled, though the graph is unecessarily complicated and visually not very attractive (for reasons below).

- I think it's wierd that it seems the author was trying to emphasize those square labels and connection lines by putting them on a higher layer over a colorful background that carries much much more information. And a more interesting thing is that, as a reader, I tend to focus on the colorful background rather than the lables on the foreground. I think this design as a defeat of purpose.

- Duplicated maps make the information carried by this image unpropotional to the ink it used. At least there should be some comments for the bottom map.

- For the color gradient triangle on the right-bottom corner, I think the author picked a horrible intensity range (agreement to (theblur05)). As it's said our eyes are more sensitive to intensity than color variations, the dimmed greyscale makes the purple part hard to read. Meanwhile the green part works well because we are able to read a wide range of spectrum of green color. I guess it's the author's intention to assign contrastive intensities to the left map and the right one, in order to show their difference. But since the two maps are already spacially separated, I think it's redundant.

- Unclear purpose of the bottom map. One thing is because the missing part. There is a isolated group of cells in the bottom map not shown in the above two maps. Another thing is the uncommented pink background circles. I would have no idea of these two things without referring to the article.

- For brevity, I agree with (theblur05)'s opinion on labels, and (nbertagnolli) and (Prashant)'s opinions on bubbles. As I see it, one obvious improvement should be replacing those small bubbles in each category with a big one. This would not only allow more white space for extra comment, but also make the image less messy. For instance, cell functions can be commented on the cluster dirrectly.
Find all posts by this user
09-07-2014, 03:51 PM (This post was last modified: 09-07-2014 03:52 PM by braden.edmunds.)
Post: #5
RE: Week 2: Slicing the wheat genome
I'm not well versed in the subject of biology, thus at first glance it was hard to understand. A few suggestions:

- The legend for both images (left and right) are not very well defined. There should be a small description explaining what each legend means. Along the same lines, I think the two images are redundant.

- The labels are giant, and should be scaled down quite a bit.

- The small "combined" image located near the bottom appears to be a waste. No information can be gathered from it (save the ink!).

- This image has a horrible color range for the dominance percentage legend. I think it would have been better to brighten the colors and possibly switch the scale to be discrete rather than continuous.

Overall, I think this image failed to convey the information effectively. If the purpose was to show a general overview of the wheat genome, it did succeed. However, for those of us who know very little about biology, it needs to be more descriptive.
Find all posts by this user
09-07-2014, 04:50 PM (This post was last modified: 09-07-2014 04:55 PM by lzhang.)
Post: #6
RE: Week 2: Slicing the wheat genome
I think theblur05 did an excellent job in describing and evaluating the visualization of homeologous gene expression of certain wheat.

  • described the type and semantics of the data and dataset;
  • described the structure and techniques used to convey information;
  • discussed the usage of color and explained why the color usage in this figure is not good;
  • discussed cons of using large label line and the distraction of the inset;
  • summarized the critique and gave suggestion on possible improvements for the visual presentation.

I want to add a few further comments.
  • I agree with theblur05's comment on color usage.
    • Some colors used in the left panel are somewhat close and not easy to discern.
    • In the right panel, the authors used color gradient to illustrate the dominance in percentage. Generally, if the data is continuous, it is a good choice to use color gradient. However, color gradient is better used when the figure itself exhibits continuous pattern, for example a surface. The right panel consists of many small circles, which are purely discrete. In this case, it is not easy to tell the subtle change of the color and the meaning of these changes. A possible improvement is that we can divide the percentage 0-100% into several ranges and use different colors to represent these ranges.
  • Regarding the labeling issue, I do not agree with theblur05. I think the figure is mainly to show the grouping structure. If this is the case, it is not an issue if some details of the figure is blocked by labels.
  • I also feel a little bit uncomfortable for the inset. For one thing, it lacks explanation. Is it the combination of the left and right figure or a third independent figure? Secondly, there is no legend explaining the color used in the inset.
  • Finally, I have some concern regarding the legend in the left figure. The width of rectangles representing the "AL" group is smaller than the other rectangles. Does this mean anything or is it purely out of aesthetic concern?

Aside from the critique, I want mention that the link given by theblur05 does not give the graph directly. Instead, this article contains the graph.
Find all posts by this user
09-07-2014, 07:33 PM
Post: #7
RE: Week 2: Slicing the wheat genome
I also agree that the labels are not a bad thing. I do agree with theblur05 that The color on the right side is sub-par and that the visualization should be interactive. I think it really depends on who the target audience is for this visualization is. Anyone can benefit from an interactive menu and this visualization confuses me. What I wonder however if to a knowledgeable person that would actually use this data such as a biologist if this would be easily understandable. This is from a magazine however so that makes me think that this is for a wider but still educated audience. I am unsure what the legends and map overview really mean. The article likely provides some context but it it is not very clear from the visualization alone.
Find all posts by this user
09-07-2014, 07:45 PM
Post: #8
RE: Week 2: Slicing the wheat genome
The visualization with the different shades of colors used and the large label looks good and simple at first glance. But as I tried to dive deep into it, it was little hard to understand with the legends and smaller inset image of the same visualization.

Few points:

1. The font of the label is readable and its clearly defines the area of the image.

2. Colors are is good on the left side but on the right side the colors are not that distinct which makes it a little hard to interpret the dominance levels it tries to show.

3. The network between the different cell types also helps to understand more about the development of the cells and the genome

4. Without theblur05's description of the inset image I would not have understood the information it is trying convey

5. The visualizations could have been made simpler by just assigning certain shapes like a circle or sphere for a category of cells instead of showing the actual cell structure which I feel makes the visualization a little cluttered.

6. I think all everyone agrees that the indexes are a disaster and at least all non-biology people would understand nothing from it especially the dominance index.
Find all posts by this user
09-08-2014, 10:51 AM
Post: #9
RE: Week 2: Slicing the wheat genome
I would mostly agree with "theblue05" on his points regarding large amount of data and small amount of information that can be deducted from graph. It took me few minutes to realize what was the intention of visualization is and there are identical visualizations on right and left. I also think the use of color gradient on the right visualization was a bad decision and its very hard to get any information out of it.
I agree with unnecessary size of category labels however I think the usage of solid white border of black lines was a good decision as not having the white border would make visualization very cluttered and black lines would get confused with network lines.
Generally I think this visualization as whole is unnecessarily cluttered and confusing.
Find all posts by this user
Thread Closed 

Forum Jump:

User(s) browsing this thread: 1 Guest(s)