Comp 790-087: Xbox Science

Dr. Phase is a game based on Dr. Mario. Your objective as “Dr. Phase” is to eliminate viruses by choosing the correct vaccinations. There are two types of viruses, red and blue. There are also two types of vaccines, red and blue, which are used to eliminate red and blue viruses respectively. If the incorrect vaccine is given to a virus it will have no effect. The smaller the virus gets, the more effect the vaccine is having in eliminating the virus. We cannot easily identify whether a virus is a red virus or a blue virus. We must see how it responds to vaccines to identify it. Yellow viruses are viruses that have not been identified yet. Viruses can transform from one type to the other depending on the vaccine being apply to them or other nearby viruses. You receive points for correctly vaccinating as many viruses as possible. You will receive bonus points for completing a level before time expires, and for maximizing the effectiveness of the vaccines (i.e. the smaller in size you can make the viruses).

The bio-informatics problem this game addresses has nothing to do with viruses or vaccines. It is actually meant to serve as a phasing application. Diploid organisms such as humans have two non-identical copies of their DNA - one they get from their mother and one they get from their father. Biologists are able to easily determine a person’s genotype, which is a string that represents both sequences of DNA combined together. This sequence tells biologists whether the two DNA sequences match or differ from each other at each position. Phasing is done to discern the individual DNA sequences from a genotype. Phasing “calls” only need to be made at heterozygous sites (places on the DNA where the two DNA sequences differ). Phasing calls are binary decisions as there are only two ways for each sequence of DNA to differ. In the game, each virus represents a heterozygous site on the DNA that needs to be phased. The vaccine you apply sets the phasing call. The color of a virus and its size are determined by a metric called ’sharing’, which essentially compares a sequence of DNA to other sequences and determines how similar they are. Each comparison “votes” for a phasing site to be called one way or the other based on the amount of sharing. The virus size is inversely proportional to amount of sharing for the given phasing call. Therefore, the smaller a virus, the better the phasing call. The hope was to learn how gamers approached the problem of phasing, to help us improve upon existing phasing algorithms or develop new ones.

The goal for this game is to fill in colors for the white boxes.  Each column represents a sequence of bases comprising a mouse’s chomosome.  Each row represents the same position in each sequence.  As a result of sequencing errors, some missing values arise and must be determined.

For each row, the player can pick from one of two colors of the total of four colors available.  Some of the missing values are truely missing (from the sequening errors), and some of the missing values were randomly inserted.  The player has control over moving the cursor, selecting colors, swapping columns, splitting groups of columns, and advancing the sequence.  Statisitics are generated based on the player’s responses to the fake unknowns in order to compute a score for the real unknowns.  After a player has finished, a scientist could use the generated statistics on the fake unknowns to deterimine possibly how good the player’s reponses to the real unknowns are.

Have a look at the “XNA refresh” info. Here’s the ReadMe, the download page, and the redistributable download.

Apparently now with “new features such as Bitmap Based Fonts, 3D Audio, Windows Vista support, Game Packaging”.

The fonts support is nice to have, as we know, and the game packaging tools supposedly make it easier to share your game.

Look at the article and then read the first post here: http://www.neowin.net/index.php?act=view&id=39747. From post, “Shame there aren’t any protein folding games out yet - it’s obvious that the PS3 was really designed for it!”

Gotta get some press for the XB.

Dan Gusfield and his student, Yufeng Wu, have taken Kyle’s challenge to improve upon his best score of 54 (which you might recall took advantage of his “cheat codes”). They have written a paper on their solution approach which will appear in the 2007 Symposium on Combinatorial Pattern Matching.

Here’s a screenshot of a 2D Prototype:

While the final version is expected to be in 3D, the upper red section represents the conveyer-belt of incoming rows of data.  The lower cyan section represents the user manipulatible section.  Rows come in from the conveyer-belt and drop onto the stack of mutable data.  Each color represents a different base (A, T, C, or G) and correspond to each of the four colored buttons on the controller.  The user can move the cursor only in the cyan section and can set values for any of bases that are unknown (grey/white).  The user can only pick from one of the two acceptable bases for that section.  Unacceptable changes are flashed with an error.

While this version is currently based on bases, a small modification could be made to convert this to the sequences of three bases to perform statistical analysis and pick from acceptable values for common/uncommon pair.

I mentioned this idea back in class last week; it is based on the Hap-Klax idea by John.  The same aspect of rows of alleles comming down a conveyer-belt is maintained; the difference arises instead from forward section.  The user instead modifies the threading of the multiple rows in the forward section to best match the haplotype/row coming down the belt.

In this way, the rows coming down the belt are generated from the source data and the manipulatible rows are like the templates.  Possible means of advancement could include sliding the data & templates to the left, either after a specified number of data haplotypes or with each one.  Points would be scored based on how well the template matches the data with the minimal number of swaps.

I’m not sure how useful this would be in this state, but there may be a useful aspect in combining this with something else.

Given a grid of data (let’s start with the easy case of binary data), permute rows & columns one at a time to form boxes. The bigger the box, the higher the score. These boxes + their underlying data show some relationships to the scientists. It lives in the same realm as the HCE and Fuzzy/MapleTree eScience apps.

So, the problem I came up with is that my sample (pseudo-random) data makes an unsolvable game given the single row/column permute rule! My house-mate and I worked on it for awhile and couldn’t form a box. What happens is that you can get really close, but when you try to move the last piece into place, it bumps something important out of the way. It’s a classic Rubik’s Cube problem, but without the degrees of freedom the R.C. allows.

Can we fix it? When Leonard presented this idea to me I instantly liked it, but we only talked about it in the end-game state.

I think it would make the data meaningless to allow sub-row/column permutations. True? Could it be solved by doing away with the single row/column at-a-time rule and allow groups of rows/columns to be permuted? Did I miss something important in the game-play concept? Is my matrix too sparse? Am I just really bad at Rubik’s Cube? I’d like to salvage this game, but I think I’m going to need the group’s ideas.

I think it’s worth further effort because you could do so many different variants. You’ve got the ho-hum 2d grid I show above, you’ve got a 3d grid that you can fly over, you can represent it as a height map (which opens up all sorts of game modes), etc, etc.

(Edit: this one is indeed up for grabs)

I don’t think I’m going to use this one, so if somebody else needs an idea…

Ok, so imagine the bastard child of HapGate and Klax…

Instead of being a tube in HapGate, you’ve got the same info coming down a conveyor as rows. You shift your catch platform right and left to line up the best match—if you can do red to red, great, otherwise do the next best matching color (my first two rows are orange to orange). Otherwise the game plays pretty much like Klax.

What’s the science you’re looking for (aside from the same sort of stuff we were talking about w/ HapGate)? I talked only briefly about this one w/ Leonard. He’s got a couple of ideas about how to make it work. If you need a game, and want to use this one, talk to him about it.

In the long run, it may or may not be viable as an eScience game.