2015.08.21 Community Chat

From Eterna Wiki

( Link back to Dev Chat archive )


On Dot Plots, Melt Plots, & Markers for High Scoring Switches


[15:01] JenPearl hi
[15:01] JenPearl This is me Jennifer PEarl
[15:01] jnicol Hi Jennifer
[15:01] JenPearl CAnt log on with Facebook
[15:01] LFP6 Assumed so
[15:01] machinelves hi Jen!!
[15:01] LFP6 Oh?
[15:02] JenPearl gives me an error
[15:02] jnicol hi elves
[15:02] LFP6 Oh, hey rainbow elf
[15:02] JenPearl This is my new alternate account
[15:02] machinelves hi hi Jnicol & LFP6
[15:02] =-= Mode #global +o jnicol by hoglahoo
[15:02] JenPearl hi Jnicol and LFP6 as well
[15:02] bekeep Hi all
[15:02] syang hi all
[15:03] JenPearl hi
[15:03] LFP6 Hey Bekeep, syang
[15:03] syang hi LFP
[15:03] machinelves hi bekeep & syang!
[15:03] LFP6 So, player chat is now, I suppose?
[15:04] machinelves yep yep
[15:04] JenPearl that is correct
[15:04] syang hi elves!
[15:04] machinelves i believe Jen is going to tell us about the wonderful mysteries of dot plots
[15:04] syang and nice new username Jennifer
[15:04] JenPearl I have a few things to share
[15:04] JenPearl thanks syang
[15:04] bekeep sweet
[15:05] JenPearl I cant log on using Facebook right now so I made a new one real quick
[15:05] machinelves i think i saw someone else mention that issue as well
[15:05] JenPearl So to start with I have not found a smoking gun yet but have found definite trends between 3 different labs
[15:07] machinelves that's great news!
[15:07] JenPearl When ploting the scores against the minimum and average pairing probabilits of the predicted structure of teh design (lower triangle of dot plot)
[15:07] JenPearl you can observe the intergediate states
[15:08] JenPearl these intermediate states can be thought of as markers for high scoring designs
[15:08] machinelves does that mean the intermediate states of pairing, as in the transitional states between one state and the next?
[15:09] machinelves is that the 'ghost' pairing that belongs to neither state one nor state two for example?
[15:09] JenPearl it is the ghost pairing
[15:09] JenPearl although sometimes teh markers can be stacks that are predictedin teh 1st state but are showing up in teh 2nd state
[15:10] JenPearl as a ghost image
[15:10] RedSpah Hi All
[15:10] syang interesting
[15:10] syang I've always wondered in labs why one part of the first state showed up in the second state sometimes
[15:11] JenPearl to go back to your first question. The intermediate states is a term that we used to describe the observation of structures inteh dot plots that should not be there compaired to single state designs
[15:11] machinelves hi Red!
[15:11] machinelves okay so not necessarily intermediate as in step between state 1 &2, just an ancillary pairing set.
[15:11] machinelves and so this is related to the trend you discovered where winning scores have an emergence of these ghost pair clusters?
[15:11] JenPearl I have found that these and other features are not really noise
[15:11] machinelves very cool
[15:11] JenPearl yes
[15:12] JenPearl the kicker is though that these designs will only show up in designs that score 60 or higher and some 80 or higher but only 50% of the top designs have them
[15:12] JenPearl so there is something else there acting on it I believe which may be melt plot stuff
[15:12] machinelves and for the set of labs that were aiming for a certain type of structure, the ghost pairs were all [ mostly ] in the same nucleotide location in the strand?
[15:13] machinelves ohh that is an interesting point. For the 50% that don't have that pattern, did they have a different pattern emerge?
[15:13] machinelves like for example, maybe they used a different sequence strategy altogether?
[15:13] JenPearl that is one thing i need to look at
[15:13] RedSpah Hi Elves, Jen
[15:14] machinelves since I only just learned about dot plots thanks to your explanation, could we have a quick intro to how to read them?
[15:14] JenPearl ok
[15:14] machinelves that way everyone can interpret these cool results
[15:14] JenPearl if you look at a dot plot there are two triangles
[15:15] JenPearl the top on is the predicted pairing probabilites for all teh nucleotide pairs and hte lower triangel has the predicted structure
[15:15] machinelves what is the easiest way to pull up a dot plot right now, through an old lab?
[15:16] JenPearl go to cloud lab and click on archive and then type in Simple RNA Switch
[15:16] RedSpah wasn't S mapped to dotplot?
[15:16] JenPearl @RedSpah Im not sure what S is
[15:17] machinelves okay found two labs, the first one?
[15:17] JenPearl scroll down to the switches lab and click on Eli Fiskers "Mod of Tebowned"
[15:18] RedSpah oh I misunderstood you
[15:18] JenPearl that is a lab i am analyzing
[15:19] machinelves cool, found it. when i click i go to the lab results, then click again i can see result for either nucleotide sequence on the left or SHAPE data on the right, which one do i want?
[15:19] JenPearl once you have that design open you can look at its dot plot
[15:19] JenPearl either really
[15:19] machinelves cool thanks!
[15:19] JenPearl but if youwant ot look at what nucs at teh sequence i would pick teh first one
[15:20] machinelves here's a direct link http://www.eternagame.org/game/solution/885046/979804/copyandview/
[15:20] JenPearl thanks
[15:21] JenPearl if you look at teh dot plot for that design there is a leg onteh uper triangle (i use leg and stack interchangabley)
[15:22] JenPearl that starts wiht nuc pair 47:42
[15:23] machinelves you can get the dot plot here: http://prntscr.com/87em6d
[15:23] JenPearl that is a marker and should not be there acording to the lower triangel in either state 1 or 2
[15:24] LFP6 Really interesting stuff Jen. I need to head off for dinner. Talk to you all later.
[15:24] machinelves bye lfp!
[15:24] JenPearl the top half uses a coordinate system of higher number nuc, lower numbered nuc
[15:24] bekeep bye lfp
[15:24] JenPearl bye LFP6
[15:24] JenPearl the lower triangel uses lower nuc then higher nuc
[15:25] JenPearl with the middle line being the pairng of the same nuc which is impossible
[15:25] machinelves hehe
[15:26] JenPearl i.e top triangle would have a pair of 42:30, teh bottom triangle would be 30:42 which is the same thing really
[15:26] JenPearl the middle would be 42:42
[15:27] machinelves http://prntscr.com/87enws
[15:27] JenPearl teh darker the blob the higher teh brobability that thart nuc pair of stack will pair
[15:27] machinelves so if you mouseover the black line by the red highlighting in that screenshot, you will see a tooltip that says the coordinates of the nucleotide positions
[15:27] JenPearl if you look at teh screenshot from Elves you can see the two red lines
[15:27] JenPearl they represent the same stack or leg
[15:27] machinelves and the black line is literally a stack or leg
[15:28] machinelves it's cool that you can see some of the shape structure in the dot plot itself!
[15:28] JenPearl thanks elves
[15:29] machinelves maybe we can get to this guy at some point, hehehe http://prntscr.com/87eonc
[15:29] JenPearl the two triangels can be thought of as mirror images of each otehr coordinate wise
[15:29] machinelves np!
[15:29] JenPearl I will let you explain that one elves
[15:30] JenPearl its your observation
[15:30] JenPearl I'll explain one more thing and then ill let elves discribe that if you ok with it?
[15:30] JenPearl @Elves
[15:30] machinelves you can actually see the loop of the aptamer site represented by that cube / circle area between the two stacks in the last screenshot. with the caveat that it's not quite a linear mapping since the two...
[15:30] machinelves ...sides of the loops are actually the top & left / bottom & right
[15:31] machinelves and Jen, I think I forgot exactly the difference between top & bottom mirrors, so why are they different again?
[15:31] JenPearl your mean the triangels
[15:31] machinelves yep
[15:32] JenPearl because RNA folds in half that iswhy the use teh top and bottom triangles. you would oonly ever be able to use the top triangle anyway so i think they reverse the cordinates so that you would not need a second chart for the structure and probs
[15:32] machinelves is bottom the structure as we designed, and the top triangle is the probability of pairing?
[15:34] JenPearl Ive observed that the bottom structure is what the predicted structure should be and the goal is to match the probabilityes with the nucs
[15:34] machinelves for example state 1 http://prntscr.com/87eqmf vs. state 2 http://prntscr.com/87eqrn
[15:34] machinelves you can see the dot plot change
[15:34] machinelves ahhh awesome thanks
[15:35] machinelves and so we were talking about a third triangle to represent real world lab results on what actually paired
[15:35] JenPearl that is becaue the state has changed and now the dot plot is showing what should happen since the energy levels have been changed by teh FMN aptamer
[15:36] machinelves so is it correct for state one that the red boxes are all the same stack? http://prntscr.com/87eriw
[15:36] JenPearl yes
[15:36] JenPearl that is corect
[15:36] machinelves and then state 2 we have a double stack with the aptamer loop in between http://prntscr.com/87ervl
[15:37] JenPearl yes
[15:38] JenPearl that is correct
[15:40] machinelves here's a visual explanation of aptamer loop http://prntscr.com/87etcc
[15:41] machinelves because it's a little tricksy - we counted out the nucleotides, and it's not exactly visually a "loop", but rather top & left or bottom & right pairings in the dot plot 'loop' between stacks
[15:41] JenPearl when we looked at a design that did not score so well that loop looked very different i remember
[15:41] machinelves yes it was all lopsided! hehehe
[15:42] machinelves so as far as the dot plot is concerned, it's represented similarly to straight lined stacks even though we see loops in our interface as curved
[15:44] JenPearl that is something I would like to analyze as well but will require som e significant programing
[15:44] machinelves cool
[15:44] machinelves hey jen is this correct? http://prntscr.com/87euvz
[15:44] machinelves keeping in mind that if we could see 2 states at once, then the bottom triangle would also have the straight line stack in black for state 1
[15:46] JenPearl blue should be state 1, green the ghost state and red is state 2
[15:46] machinelves but that middle stack in blue, is never represented in either state in the bottom triangle
[15:46] machinelves ohh thanks good eye, will correct it
[15:47] JenPearl here is interesting thing that i have observed accrois two designs of similar structure
[15:48] machinelves how about this? http://prntscr.com/87ewm2
[15:48] JenPearl that is correct
[15:48] machinelves woohoo!
[15:49] JenPearl Simple RNA SWitch and FMN Aptamer bth have a ghost pair in the middle of the first state hairpin loop inteh middle
[15:49] JenPearl FMN aptamer is one nuc off (they both have teh exact same first state and similar 2nd state
[15:50] machinelves is that this little guy? http://prntscr.com/87exa5
[15:50] JenPearl so it seams that those two similar designs like to have teh loop in the hairpin be a little bit unstable
[15:50] JenPearl yes that is it
[15:51] JenPearl that is one of the markers that shows up int eh higher scoring designs
[15:51] machinelves maybe since it is such a big loop, having a weak bond / attraction in the loop keeps the whole thing together? i have no idea, just random guess
[15:51] JenPearl I think that there are some rules we cna figure out from the dot plots
[15:52] JenPearl at this time I can only say that it seams we can be guided to a if things work out right you should be in thos range for some stuff
[15:52] machinelves it's interesting there is a huge shadow with the ghost pair section http://prntscr.com/87ey0n
[15:52] JenPearl i think so too
[15:53] machinelves and so darkest black is highest probability to pair, and lightest grey is lowest probability?
[15:53] JenPearl I have yet to analyze the light grey 'noise' but one of the modules for DPAT will be the ability to pick out those types of things. not implemented yet cause kinda hard
[15:53] JenPearl yes
[15:54] JenPearl I did some quick comarison and i would say the light grey might have a probabilty of 1-2%
[15:54] JenPearl really rough estimate based on one observation soooo
[15:55] JenPearl the singel pair in teh hairpin loop looks a little less than 9%
[15:55] machinelves which would help to understand the idea of ratio / balance between states. So if the stacks in state 1 are super dark / high prob, but stacks in state 2 are light grey, that might tell us that the switch could...
[15:55] machinelves ...potentially favor one state too much over the other, and therefore not be a flexible switch, and instead get frozen in one state. i know we have to validate that with some data, but dot plots look very useful for searching out these kinds of trends
[15:56] machinelves wow! that's good to know on the 1-2%
[15:56] JenPearl thing is that I have not found much correlation of leg probs to score just the markers. There are some observation that suggest otherwise but like I said not a smking gun yet
[15:57] machinelves and so the markers are the ghost / intermediate pairing
[15:57] JenPearl there are some ideal ranges but those almost oalways overlap with bad designs but in a smaller range
[15:57] machinelves as in, markers for scoring above 60
[15:58] JenPearl yes
[15:58] JenPearl I think that if we add melt plot that may help
[15:58] machinelves interesting!
[15:59] JenPearl I may try doing a quick manual comilation of melt points and see it that has anything
[15:59] machinelves cool!
[15:59] machinelves does anyone have any quick questions? looks like we managed to talk about dot plots for an hour already lol!!
[16:00] JenPearl did I put everyone a sleep
[16:01] machinelves personally i want to know what a melt plot is, i'm embarrassed to say i only know it has something to do with temperature at which RNA molecule falls apart. but i have never used it before
[16:01] JenPearl I havnt played with them much but it is basically just that
[16:01] machinelves i didn't even know the time had passed so count me entertained lol
[16:02] machinelves so what is the graph progression representing?
[16:02] machinelves like why isn't it a flat line? why does the line move?
[16:02] JenPearl I am actually not well versed in melt points that is my next thing
[16:03] JenPearl I remember from back in teh ealry days with single state the best melt poitnn looked light a top hat
[16:03] machinelves cool! it will be great to see how this all turns out. regardless, I'm very grateful for your research and for explaining how to read the data you're getting back, which is very interesting. thank you!!
[16:03] machinelves hahahaha
[16:03] JenPearl its 4 right now and i need to get back to work
[16:03] JenPearl np'
[16:04] machinelves so not too cold and not too hot. juuuuuust right
[16:04] JenPearl hope it helps people
[16:04] machinelves me too. take care!
[16:04] Brourd It's a prediction for the percentage of unpaired nucleotides at a specific temperature.
[16:04] syang bye Jennifer, thanks for the knowledge on dot plots
[16:04] machinelves thanks Jen! have a great day
[16:04] machinelves hi Brourd
[16:04] machinelves ohhh
[16:04] JenPearl thanks brourd
[16:04] Brourd Hello, machinelves.
[16:04] machinelves ok so 100% means 100% unpaired or falling apart?
[16:05] Brourd That would be the prediction, yes.
[16:05] machinelves so higher on the y axis = more likely for molecule to fall apart
[16:05] machinelves and further to the right of the x axis = hotter temperature
[16:05] JenPearl glad you guys like it. Talk to you all later
[16:05] bekeep Thanks for the great walkthrough
[16:06] machinelves that's funny then i would expect that the hotter it gets to always go up closer to 100% falling apart
[16:06] syang by the way, I'm just curious, did anyone take notes at eternacon?
[16:06] machinelves yeah thanks Jen! awesome help
[16:06] Brourd The higher the temperature, the more nucleotides that will be predicted to be unpaired.
[16:06] machinelves not i, syang, but good question
[16:07] machinelves okay thanks Brourd, so there aren't any freak molecules that suddenly stabilize when it gets hot lol
[16:07] syang just wondering
[16:13] Brourd Anyway, unless there are anymore questions, we'll consider this to be the end of this days player led Friday chat.
[16:14] machinelves hehehe indeed. was just posting the transcript
[16:14] machinelves my interwebz is soooo sloooowwwww
[16:15] Brourd That is an unfortunate problem to have.
[16:16] Brourd Thanks for coming, and for all those reading, please be sure to show up next Friday for the developer chat, tentatively featuring Dr. Rhiju Das of the Das lab.
[16:17] machinelves ya guess i will have to find a solution!
[16:17] machinelves here's the transcript http://eternawiki.org/wiki/index.php5/2015.08.21_Community_Chat
[16:17] machinelves thanks Brourd!
[16:17] Brourd Farewell, players.
[16:17] machinelves farewell noblefolders
[16:18] syang bye all
[16:18] machinelves bye syang! bye Brourd!