As you know, as of Monday this week, we completed the first pass of the Megathon dataset analysis, which allows us to take a peek at your answers and, as promised, provide a preliminary research result!
However, let me start with the reminder that all research is uncertain! Especially so preliminary research... 😅 So far we have completed the first pass of analysis on all movies of the dataset -- i.e. each movie has been seen at least once. We usually need more answers per movie to be 95% confident that the generated crowd answers are right, and generate the final results. But even after the first pass we have enough data to provide a preliminary result, with subsequent passes either strengthening or adjusting the preliminary findings.
So while you read the below, please keep in mind that these results are not final, and can actually change substantially after the dataset is fully complete. On the other hand, this mirrors exactly what scientists often do in the laboratory when they take an early snapshot to see if the results are moving in the expected direction. Plus, for us at the Human Computation Institute, the process of how the crowd answers reveal the research results over time raises new questions and generates new ideas regarding the crowd science as well!
What was the research question?The research question of the Megathon dataset was whether stalls occur more in the brain of mice that have high blood pressure, and whether that stalling can be reversed. If true, this could eventually lead to a preventive treatment for people that are at a higher risk of developing Alzheimer's.
Cornell's preliminary blood flow studies (using a different and less precise technique than visualizing stalls) seemed to confirm those hypotheses. To be absolutely sure though, and to look in detail at what's happening in the blood vessels under those conditions, we have to look at stalls.
So how do the results look so far?Honestly, a little wonky! :) So far, the stall analysis doesn't correlate very well with the blood flow studies that have been carried out earlier to scout out the answers to these questions. Which suggests that we will indeed need more time to generate more reliable crowd answers and refine these data.
However, it does seem from these early results that high blood pressure is associated with an increased rate of stalls in mice, and much more so in mice that have been engineered to get Alzheimer's disease. That would suggest that high blood pressure exacerbates stalling in mice already prone to get them due to the Alzheimer's gene. This could also help explain why high blood pressure is associated with a higher risk of developing Alzheimer's. But again -- that's just an early glimpse!
In the meantime, the results concerning intervention look particularly wonky. Our graphs currently show that after receiving treatment that is supposed to remove stalls, the rate of stalling actually increases in Alzheimer's mice! This doesn't add up with the early blood flow studies, so we are going to have to wait for more answers on all the movies to see what's really going on here.
That's it so far! Thanks again to everyone who contributed to the analysis thus far 💜 And please keep catching! :)Once we get enough answers, we can generate results with higher confidence and crack this nut!
This is a companion discussion topic for the original entry at https://blog.hcinst.org/megathon-prelim-research-result/