Thursday, February 14, 2008

New Pharmaceutical Sciences Flyer Artwork from Second Life

My fourth attempt at a PharmSci department flyer photo from SL.
(unanimouosly REJECTED because the colors were inappropriate).

The official photo for the 2008 PharmSci department flyer.

I am pleased to report that the faculty of the Department of Pharmaceutical Sciences approved a photo from Second Life for its official flyer! This was not easy, by any means. It actually took me five different tries over a period of a month. For the first three tries, the Second Life photo and composition was rejected after much back-and-forth discussion and argument between me and the rest of the faculty: "Colors are too bright", "The composition is too complex", "It takes too much time to figure out what is going on in this photo", "The objects do not look clear to me", "This represents you , not the department", etc. etc. But I was determined. On the third try, the biggest criticism was that the colors were still too bright. So, I emailed the faculty and told them I was going to make the colors really, really, dull. I proceeded to make the objects green and white --the colors of Michigan State, our biggest football rivals. Immediately after I emailed them the Michigan State version. I hear back from them...." the students will not like these colors" So, I immediately reworked the colors to Maize and Blue, and emailed them the photo with a text message saying "GO BLUE!"....sure enough, the photo was finally approved!

Friday, February 8, 2008

CAIA Test Successful -SL handles smoothly

I am happy to report that we have successfully uploaded 4x96 images from our styryl library dataset onto the CAIA visualization facility at ACS island in Second Life. The visualization went really well. Each image is an 8 bit black and white, 512 x 512 jpg. One is able to navigate through the images very smoothly, from my lap top computer. I am running a Dell Precision M90. Intel(r) CoreTM 2 Duo Processor T7600 (2.33GHz/667MHz/4MB) with 4GB memory. The images were displayed at full resolution and there were absolutely no glitches. Avatar navigation and camera functions were very smooth and seamless. I was later joined by my grad students CellPK Alter, CaiaLanz Alter, Kaisla Harbour, and Graziella Shostakovich, (Xinyuan, Nan, Marijo, Lilly). Diomedez Delpiaz (Prof. Juan Hinestroza) stopped by as well. It was really a lot of fun to see this running as I have been thinking about doing this for five or so years now. I must say that this really topped the Miner3D visualization software. We celebrated with virtual Bloody Mary's (which Nan and Xinyuan refused...even though the Bloody Mary's were non-alcoholic... ) with Diomedez skating around in the lab for a bit.

Sadly, from my lab's Dell quadscreen workstation, the results was not as good: when I got into the CAIA, the quadscreen practically froze. My quadscreen workstation runs the same processors and memory as my lap top, with the four 24inch LCDs powered by 2 NVidia graphics card. Marijo, one of my grad students was able to enter ACS island and navigate through the CAIA with her iBook computer no problem, so the problem was with the quadscreen and not with SL. My quadscreen is operating with Windows XP, so perhaps the problem is that I am not able to make use of the full memory and processor capabilities? My goal is to try SL on Windows Vista by the end of February, and see whether I can get any improvement in performance from the quadscreens.

Tuesday, February 5, 2008

CAIA Lab at American Chemical Society (ACS) island

Over the past week, I have been constructing a Cheminformatic Assisted Image Array (CAIA) visualization laboratory at the American Chemical Society Island in Second Life. The photo above shows the lab layout. Each side is an 8x12 array of square 2x2 meter panels, corresponding to a 96 well plate used for drug screening. Each square in the array can hold one image. In addition, each square is hyperlinked to an URL address, that can be openned by right clicking on each square.
For CAIA visualization, one can hover in the middle of the structure. Then one can get into mouselook view. From the mouselook view, it is easy to spin around with the move, and scan each side with the left and right arrows. To zoom in and out, one can use the top and bottom arrows. I gave it a try, and it is REALLY fast! Much faster than any other visualilzation software I have tried. But there is a caveat: I do not have any actual images uploaded..just the scaffold, at this point. Over the next week, I will be working with Kerby Shedden to upload some images. We are looking forward to test how the software handles with 4x96 images projected onto the scaffold.

Sunday, February 3, 2008

Avatar Entropy, Chaos, Information and Encryption in SL

Today I began some very simple experiments into the relationship between information, entropy, encryption and chaos in SL. Basically, I took three spheres (.5 m diameters) and made them so that they were kissing each other in 3D space. Then, I made a copy of these three spheres, placed them 5 meters apart from the first three spheres. I converted the 2x 3spheres into physical objects and dropped them from a distance of about 2meters from a floor plank. I repeated this experiment a few times. This is preliminary, but here is what I observed: The two sets of three spheres dropped side-by-side (simultaneously) followed similar (but perhaps not exactly identical?) paths. However, when the spheres were dropped at different times, the paths that the spheres followed upon collision with the floor plank appeared different. Caveat: this was a quick 5 minute experiment, and I simply eyeballed the results. But this is what I saw. This result brings up some interesting questions. I think the three touching spheres represents a chaotic system (someone please correct me if I am wrong). By putting 2 sets of three touching spheres side by side and dropping them simultanously, the Linden computer processors should be solving the same equations simultaneously, so the objects follow paths that are very similar to each other, upon collision with the floor plank. However, when the touching spheres are dropped at different times, the behavior of the processors may be different because the equations are not being solved simultaneously, so the paths that the spheres follow is different. Unlike a classical computer chaos experiment, it is not that the initial conditions as specified by the spatial positon of the spheres is different. Rather, the functioning of the Linden computers may be slightly different at different times. Related to this observation, one intersting question is whether 3spheres dropped SIMULTANEOUSLY at different sims from the same relative heights and relative positions would behave more similarly to each other upon collision with the floor plank, than if they werre dropped at different times. In such a case, it may be possible to encrypt information in the way in which that 3 (or more) touching spheres move upon collision with the ground, such that the information is both encoded and decoded at the point that the three three (or more) spheres are dropped from the same relative heights and positions. This encryption method would work if the physics of different sims are being processed by at the same time by the same computer. By analyzing the effect of time and distance upon differences in the paths followed by two sets of three touching spheres that are dropped from the same relative positions, it should be possible to analyze how the information content of the three touching spheres is lost as a function time and space. I will refer to this loss of information as Avatar Entropy. Nevertheless, these are preliminary observations that should be better controlled, measured and reproduced. I need to re-check that the position of the spheres are correct, that the floor planks are horizontal, etc...again these are preliminary observations. But, I want to get these results out there in case someone out there is interested in researching these questions in SL. Note: I am a pharmaceutical scientist, not an expert in information, entropy, chaos or encryption...I just think this question is sufficiently interesting that may be worth exploring. Here are a few more questions worth thinking about --- going back to a die that I was making the other day, I wonder: would two die dropped from the same position, height, and time in different sims behave the same way or differently? Would the same die dropped from the same height and position from the same sim --but at different times-- behave differently? How about the behavior of two touching spheres? How about the behavior of one sphere? How about the behavior of 1,2,3, n, n+1 kissing rods, hollow cyllinders, cubes, helices and other odd shaped objects (of the same or different geometric shapes)?

In the picture above, I am standing in front of my experimental set up, in the teaching space of Hiro Sheridan, my next door neighbor in SL. I will leave th2 2x3spheres prim behind for Hiro to take a look at. I IMed Hiro Sheridan about this. Hiro is a mathematical physicist, who works on encryption, SL, neural prosthetics, science fiction and other relevant subjects, for collaboration. Hiro will be scripting something to measure the position of the spheres where they land....measuring where the spheres land will certainly allow for a more objective assessment of my preliminary observations.

Saturday, February 2, 2008

Prim size limits, linking distance limits and the issue of scale

A while back I wrote a post about the issue of scale in SL As I have been researching in SL, I have come across two limitations of SL that bear on the issue of scale: the size limit of prims, and the distance limits of links. The size limit of prims essentially constrains any particular side of any prim to a range of 10 SL meters, to .0.01 SL meters. That means that a prim can be stretched or shrunk with the simple stretch command such that any one of its side is within 0.01 or 10 meters...not smaller, not greater. The distance link limits actually limits the distance across which different prims can be linked into a single object. How is linking and prim size limitations bear on the scale issue? Say, one wants to visualize a red blood cell, as in my previous post. I could make the red blood cell from a single prim, 10 meters maximum diameter. Or, I could make it out of cuboidal little prims -each 0.01 meters on each side-, linking them into a single object with the prim link command.

From a computer memory and speed standpoint, making the red blood cell out of a single prim is the way to go. That is the fastest and most efficient way of building the cell. But, this would mean the red blood cell is limited in size in SL, from 10 meteres to 0.01 meters. So, from ability to scale things, making them out of individual 0.01 cuboidal units would be the way to go. Except, there may be a limit to the number of prims that can be linked, and the distance over which they can be linked. Is the linking limit distance the same for small prims or for large prims? I need to find out.

Within a single prim, drug concentration can be encoded as a color or texture, but spatial variations in color or texture within the prim would not be possible. So, one would not be able to visualize concentration gradients inside the red blood cell. On the other hand, by making the red blood cell with many cuboidal 0.01 meter prims, a concentration gradient within the red blood cell could be visualized (say, for visualizing diffusion inside the cell, modeling drug transport with partial differential equations). Currently, 1cellPK models we are using do not capture intracellular concentration gradients, so visualizing cells as single (or a few) prims may be the way to go. But, if one makes the red blood cell 10 microns in diameter, than any larger cell that is scaled in relation to the red blood cell would have to be constructed from more than one prim.

In the future, if the intention is to visualize concentation gradients, it may be important to make the cells out of many single prims linked together. Actually, working with cells made out of linked prims may be unavoidable, as that is the way to visualize complex cell shapes (such as that of the red blood cell). But, there may be a limit to the number of prims that can be linked together. As things stand, this is a brief outline of the progress I have made on the scaling problem. I will proceed with more detailed visualization experiments, in terms of constraints on linking prims, and determine the interplay between number of prims, size of prims, and the distance over which prims can be linked.

Plasmodium feasting on hemoglobin (reversed time sequence)

What is the meaning of the word "visualization"? Rudolf Arnheim, Harvard University Professor, researched and wrote books about visualization and the role of visualization in guiding human thought. Short of reading these books, here is a nice, simple visuallization example: I am sitting atop a red blood cell, with a mock malaria parasite inside of it. Hemoglobin, the major protein inside the red blood cell, has a strong, red color. Through time, the malaria parasite feeds on the hemoglobin, so the hemoglobin disappears from the cell. In this visualization, the parasite is not growing, is not moving and is not reproducing. Only the hemoglobin is disappearing from the red blood cells. The top frame has less hemo than the bottom...hence I call it a reversed time sequence, because time is flowing from bottom to top frames). My avatar is about 2 microns tall, relative to the actual size of the red blood cell.
This is a very simple visualization, all done within SL. It took me less than five minutes to make this. Visualization is easier to some and harder to others, but SL is very easy, simple and fast when it comes to visuallzation work. There are no 1CellPk scripts controlling this visualization..just me manually changing the colors. So, it is just a simple illustration of visualization, meant for my graduate students to begin to understand how to use SL for 1CellPK visualizaiton purposes. In the future, I would be looking forward to my graduate students experimenting with visualizing changes occurring within the parasite inside the red blood cell. I would like to be able tosee us "treat" the red blood cell with drug (chloroquine, falcipain-2 inhibitors, etc) and visualize what happens to the drug as it distributes inside the red blood cell and then inside the parasite, and then see what it would happen inside the parasite as it is affected by the drug. Eventually, drug distributions should be controlled by mathematical models. We should be able to make simulations (5-D movies) of drugs with different chemical properties as they distribute inside the parasite. Then, we should be able to sort and analyze the dynamics observed in the movies based on drug's chemical structures. That is, Cheminformatic Assisted Image Arrays. The 5-D movies would be comprised of x,y,z spatial coordinates, signal and time. Linking the movies to chemical structures would make it 6D. Each chemical structure can be described as a multidimensional object, so in fact, a CAIA can have many, many dimensions.

About Me

I am Assistant Professor at the University of Michigan College of Pharmacy, Department of Pharmaceutical Sciences