xcms online understanding xcms algorithms · 2015. 2. 3. · h. paul benton phd the siuzdak...
TRANSCRIPT
![Page 1: XCMS Online Understanding XCMS algorithms · 2015. 2. 3. · H. Paul Benton PhD The Siuzdak Laboratory -The Scripps Research Institute. To do this morning •Learn how to fly a rocket](https://reader035.vdocuments.us/reader035/viewer/2022071103/6115ab4f5fcd99377715c122/html5/thumbnails/1.jpg)
XCMS Online
&
Understanding XCMS
algorithms
H. Paul Benton PhD
The Siuzdak Laboratory - The Scripps Research Institute
![Page 2: XCMS Online Understanding XCMS algorithms · 2015. 2. 3. · H. Paul Benton PhD The Siuzdak Laboratory -The Scripps Research Institute. To do this morning •Learn how to fly a rocket](https://reader035.vdocuments.us/reader035/viewer/2022071103/6115ab4f5fcd99377715c122/html5/thumbnails/2.jpg)
To do this morning
• Learn how to fly a rocket ship
• Get to look at chaos itself
• Shine light into the depths of nature
![Page 3: XCMS Online Understanding XCMS algorithms · 2015. 2. 3. · H. Paul Benton PhD The Siuzdak Laboratory -The Scripps Research Institute. To do this morning •Learn how to fly a rocket](https://reader035.vdocuments.us/reader035/viewer/2022071103/6115ab4f5fcd99377715c122/html5/thumbnails/3.jpg)
Ok really what are we going to do
• Learn how to fly a rocket ship– Computing hundreds of calculations at the speed of intel
– using XCMS Online
• Get to look at chaos itself– Its your data not mine !
• Data can be messy
• Chaos can being about order metabolism is highly ordered
• Shine light into the depths of nature– We get to look at some of the most complex questions
at the smallest biological level – metabolites.
![Page 4: XCMS Online Understanding XCMS algorithms · 2015. 2. 3. · H. Paul Benton PhD The Siuzdak Laboratory -The Scripps Research Institute. To do this morning •Learn how to fly a rocket](https://reader035.vdocuments.us/reader035/viewer/2022071103/6115ab4f5fcd99377715c122/html5/thumbnails/4.jpg)
Getting started with XCMSOnline
![Page 5: XCMS Online Understanding XCMS algorithms · 2015. 2. 3. · H. Paul Benton PhD The Siuzdak Laboratory -The Scripps Research Institute. To do this morning •Learn how to fly a rocket](https://reader035.vdocuments.us/reader035/viewer/2022071103/6115ab4f5fcd99377715c122/html5/thumbnails/5.jpg)
What did we do
• Registered on XCMS Online
– Confirmed real email address
• Uploaded some data
– In the old days we had to convert data ourselves –
you are all very lucky!
– XO supports – Agilent .d , Waters .RAW, Bruker .d,
AB Sciex .wiff (remember the .wiff.scan files) and
open source formats (mzML, mzXML, mzData,
netCDF)
![Page 6: XCMS Online Understanding XCMS algorithms · 2015. 2. 3. · H. Paul Benton PhD The Siuzdak Laboratory -The Scripps Research Institute. To do this morning •Learn how to fly a rocket](https://reader035.vdocuments.us/reader035/viewer/2022071103/6115ab4f5fcd99377715c122/html5/thumbnails/6.jpg)
Processing Data
![Page 7: XCMS Online Understanding XCMS algorithms · 2015. 2. 3. · H. Paul Benton PhD The Siuzdak Laboratory -The Scripps Research Institute. To do this morning •Learn how to fly a rocket](https://reader035.vdocuments.us/reader035/viewer/2022071103/6115ab4f5fcd99377715c122/html5/thumbnails/7.jpg)
Now step by step
• We’ve loaded up two datasets – 2 classes to
compare
• Set our parameters and launched a job
– Looking at the parameters and what they mean.
– Junk in, junk out. – Biologist
– Good data in, bad parameter selection, junk out –
bioinformticist
![Page 8: XCMS Online Understanding XCMS algorithms · 2015. 2. 3. · H. Paul Benton PhD The Siuzdak Laboratory -The Scripps Research Institute. To do this morning •Learn how to fly a rocket](https://reader035.vdocuments.us/reader035/viewer/2022071103/6115ab4f5fcd99377715c122/html5/thumbnails/8.jpg)
Overview of XCMS
Peak Picking
Grouping similar peaks
across replicates
Retention time
alignment
Statistical
analysis of Peaks
Between classes
![Page 9: XCMS Online Understanding XCMS algorithms · 2015. 2. 3. · H. Paul Benton PhD The Siuzdak Laboratory -The Scripps Research Institute. To do this morning •Learn how to fly a rocket](https://reader035.vdocuments.us/reader035/viewer/2022071103/6115ab4f5fcd99377715c122/html5/thumbnails/9.jpg)
Peak detection choice
Peak Picking
matchedFilter centWave
• Profile Data
• Low resolution data
• Original algorithm
• Centroid data
• High resolution data
• New published algorithm
![Page 10: XCMS Online Understanding XCMS algorithms · 2015. 2. 3. · H. Paul Benton PhD The Siuzdak Laboratory -The Scripps Research Institute. To do this morning •Learn how to fly a rocket](https://reader035.vdocuments.us/reader035/viewer/2022071103/6115ab4f5fcd99377715c122/html5/thumbnails/10.jpg)
Rockets are like ions !!
![Page 11: XCMS Online Understanding XCMS algorithms · 2015. 2. 3. · H. Paul Benton PhD The Siuzdak Laboratory -The Scripps Research Institute. To do this morning •Learn how to fly a rocket](https://reader035.vdocuments.us/reader035/viewer/2022071103/6115ab4f5fcd99377715c122/html5/thumbnails/11.jpg)
CentWave
![Page 12: XCMS Online Understanding XCMS algorithms · 2015. 2. 3. · H. Paul Benton PhD The Siuzdak Laboratory -The Scripps Research Institute. To do this morning •Learn how to fly a rocket](https://reader035.vdocuments.us/reader035/viewer/2022071103/6115ab4f5fcd99377715c122/html5/thumbnails/12.jpg)
Auto/Dynamic binning
• ROI are found by
making a first pass over
the data to find areas
that conform to
expected
chromatography and
mass spectrometer
parameters
![Page 13: XCMS Online Understanding XCMS algorithms · 2015. 2. 3. · H. Paul Benton PhD The Siuzdak Laboratory -The Scripps Research Institute. To do this morning •Learn how to fly a rocket](https://reader035.vdocuments.us/reader035/viewer/2022071103/6115ab4f5fcd99377715c122/html5/thumbnails/13.jpg)
CentWave
![Page 14: XCMS Online Understanding XCMS algorithms · 2015. 2. 3. · H. Paul Benton PhD The Siuzdak Laboratory -The Scripps Research Institute. To do this morning •Learn how to fly a rocket](https://reader035.vdocuments.us/reader035/viewer/2022071103/6115ab4f5fcd99377715c122/html5/thumbnails/14.jpg)
CentWave paramaters
• Peakwidth = How wide is your peak – from a
minimum to a maximum in seconds
• Ppm = how much does the peak vary across
scans
![Page 15: XCMS Online Understanding XCMS algorithms · 2015. 2. 3. · H. Paul Benton PhD The Siuzdak Laboratory -The Scripps Research Institute. To do this morning •Learn how to fly a rocket](https://reader035.vdocuments.us/reader035/viewer/2022071103/6115ab4f5fcd99377715c122/html5/thumbnails/15.jpg)
One thing to note
• Choose your polarity correctly!!
![Page 16: XCMS Online Understanding XCMS algorithms · 2015. 2. 3. · H. Paul Benton PhD The Siuzdak Laboratory -The Scripps Research Institute. To do this morning •Learn how to fly a rocket](https://reader035.vdocuments.us/reader035/viewer/2022071103/6115ab4f5fcd99377715c122/html5/thumbnails/16.jpg)
Retention time alignment
Obiwarp –
A Digital signal processing algorithm. Very good for high drift alignment. Fits data as if
each LC-MS 3D landscape was play dough to squeeze these together. Technically this is
warping not aligning
![Page 17: XCMS Online Understanding XCMS algorithms · 2015. 2. 3. · H. Paul Benton PhD The Siuzdak Laboratory -The Scripps Research Institute. To do this morning •Learn how to fly a rocket](https://reader035.vdocuments.us/reader035/viewer/2022071103/6115ab4f5fcd99377715c122/html5/thumbnails/17.jpg)
Retention time alignment
• Loess – this is a model to fit the data to using
the residuals to correct/align the samples
– Relies on anchors distributed across the RT
![Page 18: XCMS Online Understanding XCMS algorithms · 2015. 2. 3. · H. Paul Benton PhD The Siuzdak Laboratory -The Scripps Research Institute. To do this morning •Learn how to fly a rocket](https://reader035.vdocuments.us/reader035/viewer/2022071103/6115ab4f5fcd99377715c122/html5/thumbnails/18.jpg)
Grouping
![Page 19: XCMS Online Understanding XCMS algorithms · 2015. 2. 3. · H. Paul Benton PhD The Siuzdak Laboratory -The Scripps Research Institute. To do this morning •Learn how to fly a rocket](https://reader035.vdocuments.us/reader035/viewer/2022071103/6115ab4f5fcd99377715c122/html5/thumbnails/19.jpg)
MinFrac !
• More questions on minfrac than any other!
KO – 6 samples WT – 6 samples minFrac = 0.5 = 50%
Group become a
valid feature
Perfect biomarker
Group become a
valid feature
Just hits 50% - OK
Group is not a
valid feature
![Page 20: XCMS Online Understanding XCMS algorithms · 2015. 2. 3. · H. Paul Benton PhD The Siuzdak Laboratory -The Scripps Research Institute. To do this morning •Learn how to fly a rocket](https://reader035.vdocuments.us/reader035/viewer/2022071103/6115ab4f5fcd99377715c122/html5/thumbnails/20.jpg)
minFrac test
Not a valid
feature
A valid feature
![Page 21: XCMS Online Understanding XCMS algorithms · 2015. 2. 3. · H. Paul Benton PhD The Siuzdak Laboratory -The Scripps Research Institute. To do this morning •Learn how to fly a rocket](https://reader035.vdocuments.us/reader035/viewer/2022071103/6115ab4f5fcd99377715c122/html5/thumbnails/21.jpg)
Peak Filling
Peak not detected – intensity
filled by fillPeaks
Detected peak – peak
intensity found by
peak detector
![Page 22: XCMS Online Understanding XCMS algorithms · 2015. 2. 3. · H. Paul Benton PhD The Siuzdak Laboratory -The Scripps Research Institute. To do this morning •Learn how to fly a rocket](https://reader035.vdocuments.us/reader035/viewer/2022071103/6115ab4f5fcd99377715c122/html5/thumbnails/22.jpg)
Statistics !! Yea !!
![Page 23: XCMS Online Understanding XCMS algorithms · 2015. 2. 3. · H. Paul Benton PhD The Siuzdak Laboratory -The Scripps Research Institute. To do this morning •Learn how to fly a rocket](https://reader035.vdocuments.us/reader035/viewer/2022071103/6115ab4f5fcd99377715c122/html5/thumbnails/23.jpg)
Adduct selection
![Page 24: XCMS Online Understanding XCMS algorithms · 2015. 2. 3. · H. Paul Benton PhD The Siuzdak Laboratory -The Scripps Research Institute. To do this morning •Learn how to fly a rocket](https://reader035.vdocuments.us/reader035/viewer/2022071103/6115ab4f5fcd99377715c122/html5/thumbnails/24.jpg)
![Page 25: XCMS Online Understanding XCMS algorithms · 2015. 2. 3. · H. Paul Benton PhD The Siuzdak Laboratory -The Scripps Research Institute. To do this morning •Learn how to fly a rocket](https://reader035.vdocuments.us/reader035/viewer/2022071103/6115ab4f5fcd99377715c122/html5/thumbnails/25.jpg)
Cloud plot
Size = fold change
Colour = signficance (lower
p-value)
Black or white ring = metlin
hits
![Page 26: XCMS Online Understanding XCMS algorithms · 2015. 2. 3. · H. Paul Benton PhD The Siuzdak Laboratory -The Scripps Research Institute. To do this morning •Learn how to fly a rocket](https://reader035.vdocuments.us/reader035/viewer/2022071103/6115ab4f5fcd99377715c122/html5/thumbnails/26.jpg)
Static PCA
![Page 27: XCMS Online Understanding XCMS algorithms · 2015. 2. 3. · H. Paul Benton PhD The Siuzdak Laboratory -The Scripps Research Institute. To do this morning •Learn how to fly a rocket](https://reader035.vdocuments.us/reader035/viewer/2022071103/6115ab4f5fcd99377715c122/html5/thumbnails/27.jpg)
![Page 28: XCMS Online Understanding XCMS algorithms · 2015. 2. 3. · H. Paul Benton PhD The Siuzdak Laboratory -The Scripps Research Institute. To do this morning •Learn how to fly a rocket](https://reader035.vdocuments.us/reader035/viewer/2022071103/6115ab4f5fcd99377715c122/html5/thumbnails/28.jpg)
Results.zip download file
• This has all of the plots and information from
the processed job.
– Static PCA
– Static heat map
– Static cloud plots
– Scaling plot – Good for looking at scaling for PCA
(trend implicates heteroscedastic noise)
![Page 29: XCMS Online Understanding XCMS algorithms · 2015. 2. 3. · H. Paul Benton PhD The Siuzdak Laboratory -The Scripps Research Institute. To do this morning •Learn how to fly a rocket](https://reader035.vdocuments.us/reader035/viewer/2022071103/6115ab4f5fcd99377715c122/html5/thumbnails/29.jpg)
Contents of results.zip file
XCMS.diffreport. And XCMS.annotated.diffreport are the data tables with all the
intensity values associated with them not results.tsv
![Page 30: XCMS Online Understanding XCMS algorithms · 2015. 2. 3. · H. Paul Benton PhD The Siuzdak Laboratory -The Scripps Research Institute. To do this morning •Learn how to fly a rocket](https://reader035.vdocuments.us/reader035/viewer/2022071103/6115ab4f5fcd99377715c122/html5/thumbnails/30.jpg)
Thank you
Prof. Gary Siuzdak Duane Rinehart
Dr. Bill Webb
Questions?
Comments? Thoughts?
![Page 31: XCMS Online Understanding XCMS algorithms · 2015. 2. 3. · H. Paul Benton PhD The Siuzdak Laboratory -The Scripps Research Institute. To do this morning •Learn how to fly a rocket](https://reader035.vdocuments.us/reader035/viewer/2022071103/6115ab4f5fcd99377715c122/html5/thumbnails/31.jpg)
Obiwarp details