PolySNAP 3 intro

1
(No Transcript)
2
Introducing PolySNAP 3

• PolySNAP 3 is the successor product to the
  groundbreaking PolySNAP
• PolySNAP provided powerful cluster analysis tools
  and clean, elegant and informative graphic
  displays to allow for high-throughput analysis of
  Powder X-Ray Diffraction patterns.
• PolySNAP 3 has the same powerful tools, but
  allows them to be employed to ANY kind of numeric
  data
• PXRD, Raman, IR, DSC, Melting Points. Anything!
• PolySNAP 3 allows up to four different datasets
  to be analysed simultaneously
• (or a single dataset with up to four different
  processing options)
• PolySNAP 3 combines the results from individual
  datasets to show the overall picture of whats
  going on

3
Multiple Datasets

• Combined XRPD Raman instruments now available
  D8 Spectrolab
• Applying multiple techniques to the same samples
  helps give additional information to work with
• How would we actually combine results from two
  such different techniques ?
• Automatically calculate optimal weighting for
  each entry in each dataset
• Much more powerful than manually averaging
  results

4
Methodology- PXRD Raman
Full profile matching all patterns against all
patterns
n XRPD Patterns
nxn Correlation Matrix
XRD results
nxn Distance Matrix
Combined results
nxn Distance Matrix
Combine
Full profile matching all patterns against all
patterns
n Raman Patterns
nxn Correlation Matrix
nxn Distance Matrix
Raman results
5
(No Transcript)
6
(No Transcript)
7
Run ona single dataset

• Just like PolySNAP
• - but can now select
  different kinds of input
• - Separate automatic
  pre-processing defaults for each type of
  data
• - also new is the
  ability to transform spectra
• - either Fourier
  transform or derivative (1st or 2nd)
• - can also specify a
  location to load sample image files from
  (JPEG files)

8
Analysis ToolsDendrogram

• Similar patterns are clustered together
• Position of Cut-Level partitions the data into
  separate clusters
• Adjust the cut-level to best
• describe the data

Too high?
Too low?
Just right
9
Analysis Tools3D MMDS Plot

• Each sphere is a pattern
• The closer together two spheres are, the more
  similar they are
• Similar patterns clump together
• Different patterns are far apart
• Easy to spot outliers at a glance
• As with all PolySNAP graphics, you can
• Zoom in, zoom out, rotate
• Hide and show and drag labels
• Change rendering quality
• Change colour of background, axes etc.
• Produce figures for publication

10
Multiple MethodsConsistent Colouring

• Changing the clustering in the dendrogram
  automatically updates the colours in the other
  plots.

• Cluster colours from dendrogram are used on MMDS
  plot so you can compare the groupings from both
  methods. Same results from both increases
  confidence in the accuracy of the results.

11
Analysis ToolsQuantitative Analysis

• Given reference phases, PolySNAP can identify
  possible mixtures and perform fast quantitative
  analysis. Results are shown on the Cell Display
  in the form of Pie Charts or Stacks.

12
Analysis ToolsCell Display

• Cell Display is also useful for seeing the
  clustering result colours superimposed on e.g.
  the layout of an original 96 well plate from a
  high-throughput system.

13
Analysis Tools6D Plots Sample Preparation
Information

• Automatically read sample preparation information
  from data file headers or standalone CSV file
• Plot any combination of this information as
  different variables in the 3D plot vary size,
  shape and colour of plotted points
• e.g. show different solvents as different
• shapes
• Vary the shape size with temperature,
• the colour with time
• See what initial variables correlate to
• resulting materials

14
Analysis ToolsValidation

• Check the validity of the clustering using
  powerful statistical tests
• Silhouettes are these cluster memberships
  reasonable?
• Fuzzy Clustering should this pattern be in a
  different cluster?
• Scree plots how many clusters are needed to
  explain the data?
• Minimum Spanning Trees an alternative way to
  construct clusters
• Silhouettes example
• Pattern 21 scores poorly as a member of this
  cluster
• it is the least tightly connected in the
  dendrogram
• This turns out to be a mixture, rather than a
  pure phase

15
Analysis ToolsHigh Dimensionality Viewers

• Advanced visualisers to test if the clustering
  holds true in higher dimensional space
• Animated dataset Grand Tour
• Parallel Coordinates Plots
• Space Explorer View

16
Analysis ToolsPattern Thumbnail Viewer

• Thumbnail colouring updated as dendrogram cut
  level is changed
• Helps show which samples are grouped together in
  which clusters
• Useful visual overview of smaller datasets

17
Analysis ToolsColour-coded Numeric Results

• Draws the eye to help see patterns in the data
• User-controlled colour scheme and cut-off points

18
Analysis ToolsPeak Overlap Viewer

• Overview of peak intensities and positions
  throughout the entire dataset
• Colouring can be user controlled
• See at a glance if peaks coincide or overlap at
  similar angles

19
Run onMultiple Datasets

• 3 potential ways of working with PolySNAP
• 1. Multiple types of data have been collected on
  the same samples
• e.g. there is a set of 96 samples, from which
  have been collected both PXRD and Raman spectra
• Examine the clustering from the PXRD, and the
  Raman, and the combined PXRD Raman
• Do the different methods agree? Do they
  contradict each other?
• 2. Multiple instances of the same type of data
  collected
• e.g. PXRD patterns collected on the same samples
  at different times, or under different conditions
• 3. Investigate a single dataset under different
  processing options
• Compare side-by-side what difference turning
  background subtraction, etc. on makes to the
  final results

20
Run onMultiple Datasets
Select PXRD and choose data location
Select Raman and choose data location
Select any different processing options for each
dataset
Click OK to start
21
How much data?

• For each individual dataset, PolySNAP can analyse
  up to 1,500 samples at once
• Up to four different datasets can be input for
  each run
• The datasets can be any combination of
• PXRD
• Raman
• DSC
• IR
• Other spectra data
• Numeric
• Numeric can be a correlation matrix, or raw
  numeric data relating to the samples being
  analysed
• e.g. Melting points
• Data formats RAW, Powder-CIF, ASCII, CSV, Raman
  SPC, Opus,

22
Results Screen
23
Results Screen
Click on the dataset name to switch between
different sets of results here, choose between
PXRD and Raman individually, or the combined
results
24
Results Screen
Click on the tabs to see different display
screens to help you interpret the results
25
Results Screen
Select a sample or samples in the results screen
to see information about it and its profile
displayed below
Click the vertical tabs to switch between seeing
the PXRD or Raman profile for the currently
selected patterns
26
Results Screen

• Compare the clustering results from each datatype
  in turn

27
Results Screen

• Look at both the PXRD and Raman spectra to make
  sure this clustering looks sensible

28
Results Screen
Or select up to four multiple datasets to view
results side by side
29
Multiple ScreensMultiple Windows

• Customers with multiple monitors can open
  separate results screens on each one, allowing
  easy comparison of different results side-by-side

30
Combined DatasetsExample 1

• 48 patterns of 3 forms of Sulfathiazol (forms 2,
  3 and 4)
• PXRD and Raman data collected
• PXRD Data only
• Splits form 3 into 2 separate clusters

Form 2
Form 3
Form 3
Form 4
31
Combined DatasetsExample 1

• 48 patterns of 3 forms of Sulfathiazol (forms 2,
  3 and 4)
• PXRD and Raman data collected
• Raman Data only
• Doesnt distinguish between Form 3 and Form 4

Form 2
Forms 3 4
32
Combined DatasetsExample 1

• 48 patterns of 3 forms of Sulfathiazol (forms 2,
  3 and 4)
• PXRD and Raman data collected
• Combined PXRD Raman using Automatic Weights
• Does much better than the individual methods alone

Form 3
Form 4
Form 2
33
Combine ResultsExample 2

• 46 patterns of 2 anhydrous forms of
  Carbamazepeine (Forms 1 3)
• PXRD and Raman data collected
• PXRD Data only
• E3 and F7 in different clusters

34
Combine ResultsExample 2

• 46 patterns of 2 anhydrous forms of
  Carbamazepeine (Forms 1 3)
• PXRD and Raman data collected
• Raman Data only
• E3 and F7 in same cluster!

35
Combine ResultsExample 2

• 46 patterns of 2 anhydrous forms of
  Carbamazepeine (Forms 1 3)
• PXRD and Raman data collected
• PXRDRaman Data Combined
• F7 highlighted as an outlier due to this
  inconsistency
• 
  Other outliers (yellow) are
  mixtures of the 2 forms

36
Combined Results

• Matching method does very well in distinguishing
  forms automatically using either Raman or PXRD
  data
• Combined results using Automatic Weights seem to
  do better than either PXRD or Raman individually
• Identification of pure phases / mixtures improved
• Use of combined data highlights any
  inconsistencies in separate analyses
• Such inconsistencies would not be obvious with
  only one data source
• User can then examine outliers manually in detail
• Seeing similar clustering from multiple original
  data sources increases confidence in the results

37
Quality Control

• Given a set of reference patterns, new patterns
  can be considered to be similar enough to the
  references to pass, or different enough to
  fail.
• Graphical representation
• new samples within the green
• Pass surface are OK, samples
• falling outside the surface fail.

38
Pre-screening

• Have a large dataset of existing patterns
  (100,000) ?
• Compare a single new unknown sample to the large
  dataset in minutes
• Get back a list of top 50 best matches to the
  unknown
• Perform PolySNAP clustering visualisation on
  the best matches
• Pre-screen a large dataset down to a smaller,
  more relevant subset

39
Other Features

• Automatic Report Writer
• Ability to run from command line - perform
  analysis as part of a script
• Integration with other Bruker software
• Audit-trail detailed logfile
• Raw numeric results available for interrogation
• Results automatically archived to encrypted file
• Full manual and tutorial provided
• View sample images alongside patterns
• Manual analysis mode
• Compare one pattern to many
• Detailed numeric breakdown of results

40
Dealing withLarge Datasets

• Specially designed tools for dealing with large,
  complex datasets
• Select subsets for re-analysis
• View simplified dendrograms and 3D plots
• Show/Hide selected clusters
• Hide all but current cluster being investigated
• Transparency options
• Search options locate individual patterns of
  interest

41
PolySNAP M

• Ideal for single, smaller datasets (up to 96
  patterns)
• PolySNAP M has dendrograms, 3D plots,
  prescreening and manual analysis mode

42
Which PolySNAPDo I Need?
PolySNAP M PolySNAP 3
Program Type Commercial Commercial
Input Data Types Powder XRD PXRD, Raman, DSC, IR, Numeric, Other
Max No. Of Samples per run 96 4x 1500
6D plots of sample prep info No Yes
Sample JPEG display No Yes
Cluster Validation Tools No Yes
Report Writer No Yes
Re-run selected analysis No Yes
No. of clustering methods 1 5
Online documentation Yes Yes
Program Tutorial Yes Yes
Customisable Analysis Methods No Yes
Manual Matching Options Yes Yes
Interactive Dendrograms Yes Yes
3D MMDS Plots Yes Yes
3D PCA Plots Yes Yes
Numerical Results Yes Yes
Audit-trail Logfile Yes Yes
Amorphous Identification Yes Yes
Reference patterns for mixtures Yes Yes
Command-line run options No Yes
Quality Control mode No Yes
Combine multiple datasets No Yes
Multiple Monitor Support No Yes
43
Support Services

• Training
• One and two-day in-depth training packages
  available
• Gain detailed knowledge of advanced program
  functions
• Use advanced options with trickier datasets with
  confidence
• Look at your own data with
• Customisation
• Personalised modifications to suit
  company-specific workflows
• Interface alterations
• Integration with other custom-software

44

www.bruker-axs.com
www.chem.gla.ac.uk/snap