Amazing video. Thank you so much. I would like to know how you'd show the difference in certain cell population between lets say a control vs patient samples? The overlay is fine however the clusters arent clear on that overlay map. Again thanks!
Oh my, I lately have a lot of these "bad luck FlowJo days". So I totally get you. Everytime that I try to concatenate my populations, FlowJo crashes. Same thing happens when I use Cluster Explorer...It's so frustrating. Btw, love your channel🙂
excellent video, Unfortunately i'm having problems with t-sne. At keep ending with a super huge single cluster, there is no bunderies to form new cluester even i selected different markers. Don't know what i'm doing wrong
Hi! Sorry I missed this question. I'm not actually sure what would cause this- could be your setting or a homogenous population. You could try running another dimensionality reduction (such as UMAP) to see if that works better for you
Hi! Thanks again for such a great video. I was wondering if there was a way to set thresholds for the clusters. I understand that clusters are generated based on relative expression when compared to other clusters (rather than absolute FMIs) but when I last generated clusters with FlowSOM on my data, it appeared that I had 2 clusters with high expression of Ly6G while in reality the detection of fluorescence in Ly6G stems probably from high AF or spillover from other fluorophores (my samples are PMBCs obtained after Ficoll-gradient generated separation; this was also confirmed by doing a Ly6G FMO). Would there be a way for accounting for things like this when generating clusters?
It's recommended that you run clustering on compensated data so things like spillover aren't affecting your results. The algorithm is simply using the signals you give it, so they must be as clean and clear as possible (so compensated and cleaned using something like FlowAI or FlowClean and normalized if necessary). If one of the Ly6G populations seems to be due to autofluorescence, I would recommend going back and taking a look at your panel and redesigning so that this marker is not falling into the area where you have autofluorescence (or run on a spectral analyzer if you have access and subtract out AF).
thank you so much!! I have a question regarding the keywords - I used "SampleID" to distribute my numerical register for my samples, but for some reason it never shows up correctly. When using "cell source" just as you do it seems fine. Is there any explanation behind this? Also, FlowSom keeps on failing when running it (without any error code), are there any tipps and tricks? Thank you!
Weird- I've not seen that happen before. If cell source is working you should be fine. Unsure why FlowSOM keeps failing. I've found sometimes reinstalling the plugin can help. And make sure you are using the correct version of R
Hi Aja, Thank you so much for making these videos! They have been super helpful. I do have a couple of questions and would love it if you could help me with them. First, I noticed that you had gated all of your populations prior to concatenating the live CD45+. Do the downstream gates somehow translate over for tSNE analysis? I just gated up to CD45+ and then concatenated the files (4 groups in total, numbered 1-4), which leads me to my next question. After I run the tSNE and try to use the Cluster Explorer, I don't seem to get the pop-ups like I see in your video. So, I was wondering if it was because of the pop-ups (and I need to figure out why they aren't enabled), the fact that I'm using a samsung and I see you're working with a mac, or because of the gating tree you had before concatenating. I hope this makes sense. Any insights would be helpful. Gratefully, B.
Hi, I've already thanked you in the past for the amazingly clear and helpful video. I am going to renew this again but also ask a question: when running FlowSOM, how do you decide on the number of clusters to identify? It seems to me that deciding subjectively is a little unstandardised, but I could not find any real guidelines. Also, is there a way to go below 3 clusters? Every time I input "2" as number of clusters, I would not get a result. As in, it looks like it is running, but no graphs will ever appear. Many thanks
Thanks and I'm glad the videos have been useful to you! I think you are not getting a result because you are "underclustering". While there is no consensus on how to subjectively decide the number of clusters (at least not one I could find either), the main consensus seems to be that it is better to over estimate clusters than under estimate. Take a look at this publication (f1000research.com/articles/7-1141#f3)- it gives a good look at the clustering algos out there. In their work, they found that FlowSom "required a relatively large number of clusters to reach its maximal performance". CytoBank has a pretty good description of how to go about making this decision (support.cytobank.org/hc/en-us/articles/360015918512-How-to-Configure-and-Run-a-FlowSOM-Analysis)- scroll down to the "Choosing a target number of clusters". If all else fails, maybe try the "Auto" function to see if you at least get something! I hope that helps!!
Hi Aja, thanks heaps for the video. What version of flowjo are you using? My version doesn't seem to allow the multigraph colour mapping only histograms, all parameters by Y and NXN plots.
Hi Emily- glad it was helpful! I have just updated to the newest version (10.8.1). If I remember correctly, I believe I was on 10.7 when this was filmed. Always looking for ideas for new videos- so if there are any topics that would be beneficial for you, let me know!
Hi Aja, I've been following your videos for a while, can you recommend any online classes to get a certificate for flow cytometry? I'm a MT that recently was transferred to the flow cytometry department and I would like to get some education related to it. Thank you
Hi Raymir. This is something that I offer (send me an email at aja@ualberta.ca). You can also check out the ExCyte program- I believe they offer a certificate upon completion as well.
Hi, thanks for this great video. Can I use tSNE to analyse, let say, 2 groups of 5 mice each? should I concatenate the 5 mice from each group into a file, therefore get 2 concatenated files and tSNE them? Thanks
It all depends on your questions. You could also sign multiple keywords to each sample (one for mouse # and one for group) and make a single concatenated file that you run the tSNE on.
little issue: I added keyword "tSNE_group'' to all 10 FCS files (10 mice) (with value bing 1 or 2), but the concatenated file does not show the keyword, weird...
@@vblanche1 Sorry- this got lost in the shuffle! To get this included, go into the advanced part when you are on the concatenation page and into the keywords section. In here, you can select the keywords you added so they get included in the concatenated file.
![FlowJo [STAIN INDEX]](http://i.ytimg.com/vi/R-BfoEoDNXQ/mqdefault.jpg)
![FlowJo [STAIN INDEX]](/img/tr.png)
One of the best tutorials I have even seen on TSNE. Thanks so much!
You are the best teacher on Flow. Thank you so much 😊😊
I really loved your tutorial!!! Also because the program crashed while doing it, showing real life situation! Thanks a lot!
Thanks!
Aja, thanks, this is super useful!
Thanks! More coming soon- any topics you would like covered???
I really loved your tutorial. Thank you very much.
Thanks!
Yess, finally I get what all this hustle about Tsne plots. And Aja, OMG bad luck is realllll
Amazing video. Thank you so much. I would like to know how you'd show the difference in certain cell population between lets say a control vs patient samples? The overlay is fine however the clusters arent clear on that overlay map. Again thanks!
Oh my, I lately have a lot of these "bad luck FlowJo days". So I totally get you. Everytime that I try to concatenate my populations, FlowJo crashes. Same thing happens when I use Cluster Explorer...It's so frustrating. Btw, love your channel🙂
Thanks! And good to know I'm not the only one who gets hits with the bad luck FlowJo days 🙃
such an amazing channel！
Thanks 😊
Thank you so much! Could you do one on how to get UMAP plots, too?
I sure can! I’ll film that this week
@@ajarieger_flow thank you so much! And great work by the way! So clear, so useful, so very well done !!
@@user-rh8oe2vj4c UMAP video is out:
th-cam.com/video/fuTOiaPr86o/w-d-xo.html
excellent video, Unfortunately i'm having problems with t-sne. At keep ending with a super huge single cluster, there is no bunderies to form new cluester even i selected different markers. Don't know what i'm doing wrong
Hi! Sorry I missed this question. I'm not actually sure what would cause this- could be your setting or a homogenous population. You could try running another dimensionality reduction (such as UMAP) to see if that works better for you
Finally 🤗🤗🤗
Hi! Thanks again for such a great video. I was wondering if there was a way to set thresholds for the clusters. I understand that clusters are generated based on relative expression when compared to other clusters (rather than absolute FMIs) but when I last generated clusters with FlowSOM on my data, it appeared that I had 2 clusters with high expression of Ly6G while in reality the detection of fluorescence in Ly6G stems probably from high AF or spillover from other fluorophores (my samples are PMBCs obtained after Ficoll-gradient generated separation; this was also confirmed by doing a Ly6G FMO). Would there be a way for accounting for things like this when generating clusters?
It's recommended that you run clustering on compensated data so things like spillover aren't affecting your results. The algorithm is simply using the signals you give it, so they must be as clean and clear as possible (so compensated and cleaned using something like FlowAI or FlowClean and normalized if necessary). If one of the Ly6G populations seems to be due to autofluorescence, I would recommend going back and taking a look at your panel and redesigning so that this marker is not falling into the area where you have autofluorescence (or run on a spectral analyzer if you have access and subtract out AF).
thank you so much!! I have a question regarding the keywords - I used "SampleID" to distribute my numerical register for my samples, but for some reason it never shows up correctly. When using "cell source" just as you do it seems fine. Is there any explanation behind this? Also, FlowSom keeps on failing when running it (without any error code), are there any tipps and tricks? Thank you!
Weird- I've not seen that happen before. If cell source is working you should be fine. Unsure why FlowSOM keeps failing. I've found sometimes reinstalling the plugin can help. And make sure you are using the correct version of R
Hi Aja,
Thank you so much for making these videos! They have been super helpful. I do have a couple of questions and would love it if you could help me with them. First, I noticed that you had gated all of your populations prior to concatenating the live CD45+. Do the downstream gates somehow translate over for tSNE analysis? I just gated up to CD45+ and then concatenated the files (4 groups in total, numbered 1-4), which leads me to my next question. After I run the tSNE and try to use the Cluster Explorer, I don't seem to get the pop-ups like I see in your video. So, I was wondering if it was because of the pop-ups (and I need to figure out why they aren't enabled), the fact that I'm using a samsung and I see you're working with a mac, or because of the gating tree you had before concatenating. I hope this makes sense. Any insights would be helpful.
Gratefully,
B.
Hi! Thanks for reaching out! It's hard to know without seeing your data... feel free to send me an email at aja@ualberta.ca and we can discuss further
Hi, I've already thanked you in the past for the amazingly clear and helpful video. I am going to renew this again but also ask a question: when running FlowSOM, how do you decide on the number of clusters to identify? It seems to me that deciding subjectively is a little unstandardised, but I could not find any real guidelines. Also, is there a way to go below 3 clusters? Every time I input "2" as number of clusters, I would not get a result. As in, it looks like it is running, but no graphs will ever appear. Many thanks
Thanks and I'm glad the videos have been useful to you! I think you are not getting a result because you are "underclustering". While there is no consensus on how to subjectively decide the number of clusters (at least not one I could find either), the main consensus seems to be that it is better to over estimate clusters than under estimate. Take a look at this publication (f1000research.com/articles/7-1141#f3)- it gives a good look at the clustering algos out there. In their work, they found that FlowSom "required a relatively large number of clusters to reach its maximal performance". CytoBank has a pretty good description of how to go about making this decision (support.cytobank.org/hc/en-us/articles/360015918512-How-to-Configure-and-Run-a-FlowSOM-Analysis)- scroll down to the "Choosing a target number of clusters". If all else fails, maybe try the "Auto" function to see if you at least get something!
I hope that helps!!
Hi Aja, thanks heaps for the video. What version of flowjo are you using? My version doesn't seem to allow the multigraph colour mapping only histograms, all parameters by Y and NXN plots.
Hi Emily- glad it was helpful! I have just updated to the newest version (10.8.1). If I remember correctly, I believe I was on 10.7 when this was filmed. Always looking for ideas for new videos- so if there are any topics that would be beneficial for you, let me know!
Hi Aja, I've been following your videos for a while, can you recommend any online classes to get a certificate for flow cytometry? I'm a MT that recently was transferred to the flow cytometry department and I would like to get some education related to it. Thank you
Hi Raymir. This is something that I offer (send me an email at aja@ualberta.ca). You can also check out the ExCyte program- I believe they offer a certificate upon completion as well.
Hi, thanks for this great video. Can I use tSNE to analyse, let say, 2 groups of 5 mice each? should I concatenate the 5 mice from each group into a file, therefore get 2 concatenated files and tSNE them? Thanks
It all depends on your questions. You could also sign multiple keywords to each sample (one for mouse # and one for group) and make a single concatenated file that you run the tSNE on.
@@ajarieger_flow thanks for your suggestions. my idea is to compare group 1 against group2 , group1 untreated mice, group2 treated mice.
little issue: I added keyword "tSNE_group'' to all 10 FCS files (10 mice) (with value bing 1 or 2), but the concatenated file does not show the keyword, weird...
@@vblanche1 Sorry- this got lost in the shuffle! To get this included, go into the advanced part when you are on the concatenation page and into the keywords section. In here, you can select the keywords you added so they get included in the concatenated file.