Our World Statistics Day conversations have been a great reminder of how much statistics can inform our lives. Do you have an example of how statistics has made a difference in your life? Share your story with the Community!
Choose Language Hide Translation Bar

SIFTomics and Data Analytics: The Quickest Way to Unravel the Totality of Your Chemical Space (2020-EU-30MP-414)

Level: Intermediate


Camilla Liscio, Senior Application Chemist, Anatune
Jamie Minaeian, Application Chemist, Anatune


In a world of increasing complexity, analytical chemists must unravel the entirety of the chemical space of products and materials. On this never-ending quest from complexity to clarity, data analytics becomes an essential tool. VOCs are known to impart an odor to products. The traditional approach to quantifying odor uses a sensory panel, which is expensive and can be subject to problems brought about by fatigue. Selected Ion Flow Tube Mass Spectrometry (SIFT-MS), however, can selectively detect and quantify a wide range of odor compounds in real time, more cost-effectively. The challenge is how to make sense of the rich data set generated by fast SIFT-MS analysis.This is where JMP machine learning and multivariate analytics brings clarity by enabling extraction and understanding of the most important chemical insight. This talk will demonstrate the synergic power of SIFT-MS analysis combined with chemometrics to characterize the chemical space of odor compounds in a real application scenario.



Wonderful presentation, @CamillaLiscio.


There was no need to thank me. You did all the work!


Have you tried using the “smell scale” (e.g. 1-6) as a response rather than just the sample classifications. I think that could be interesting and maybe uncover some more insights. I can't remember if we talked about this. But I would be happy to help you try this.


I'm thinking a parallel plot coloured by the green/amber/red scale that you have could be really powerful!

two comments:

1. you were lucky to get 22 samples covering the full range. where they selected from a larger set?

2. the use of covariates reflecting the production process would have been very informative. do you have such data?

Hi Camilla, this question came in during your session: Question to Camilla: What made her select the PCA platform, straight after the Predictor Analysis? Could have she used the Graph Builder as an intermediate step to get some initial insight?



Thank you Phil! You definitely helped us getting started though! It is a very good idea, we will be processing further datasets so I think it would be good to try that. Will speak soon


Thanks for your questions.


1. They were chosen from a larger set of samples but there were several practical reasons why this pilot study was planned this way, one of the main drivers being the sensory score data

2. Not at the moment, I guess that will be probably one of the next steps



One of the advantages of using PCA is reducing the dimensionality of a multivariate space and PCA seemed to be a very straightforward way to visually and effectively explore the results also considering we were hoping to move to a classification modelling option afterwards

Hi Camilla, great & very useful presentation!

I wonder if Partial Least Squares (PLS) may not be an additional effective tool for quantifying the relationship between odor components and spectral MS data as is also done with NIR?

@frankderuyck Thank you Frank! We are still finding our feet with this project and we very much welcome any suggestion. If you have any advice or recommendation to share based on your experience please do reach out to us!