[Paul Hiscoe]

Hi Giorgio

 

We are very interested in this. What Model fitness R2/RMSE did you manage to achieve? Do you have a verified set of testing results?

 

Thanks

 

 

Paul

[Paul Hiscoe]

Hi Matteo

 

We have now released an Android app that lets you store the results of your tests in the cloud

 

https://play.google.com/store/apps/details?id=com.ph7.analyserforscio

 

Paul

[Paul Hiscoe]

So as far as we can tell gluten is actually a complex of two groups of proteins.
The two groups are called gliadins and glutenins. Each group contains lots of different proteins which all have a similar structure.
These make up around 80% of the storage proteins found in the flour. Storage proteins are used by the plant to store things like iron and amino acids.
When in the flour they create a network by linking together – the two proteins together are what is called gluten.

When manufacturers extract gluten they run dough under water which removes most of the starch and other material in the cells. The mixture is then dried to give a flour like substance. This however only gives around 75% gluten (protein). The majority of what is left is starch (a carbohydrate), but there are also lipids (fats) and some other proteins will still be present. So how to explain our ‘100% gluten’ packaging which lists the contents as 78% protein?

 

If we can figure out what percentage gluten rather than protein in any given product (examples of zero are easy) I think we might be able to produce a model.

 

Has anyone got any insight as to how to determine the gluten% ?

[Paul Hiscoe]

We’re currently working on this area. We’ve just come back from Australia where we have acquired a sample claiming to be 100% Gluten flour (even though it is only 89% Protean).

 

Initial tests show that we are able to estimate the overall level of protean in a variety of wheat flours. Flours claiming to be gluten free  are, of course, not made from wheat. We’re going to try adding them into the collection and see if we can build a model to distinguish. Any scientific input on this most welcome.

[Paul Hiscoe]

 

We’re using a liquid accessory and getting R2 = 0.997 | RMSE = 0.642 for our Vodka model.

[Paul Hiscoe]

It works very well.

 

We’ve now built an ethanol model with R2 = 0.997 | RMSE = 0.642

 

We’ve tested it extensively and had the results verified by the Public Analyst.

[Paul Hiscoe]

Did you switch outlier detection on?

 

[Paul Hiscoe]

We’ve developed a model that is successful in measuring Alcohol% in Vodka and are interested in extending this work to beer.

 

If you are interested in deploying a beer model, please contact me.

 

Paul

[Paul Hiscoe]

PFB:

 

I bought some PTFE sheet and made my own cup, but it’s a bit crude. I would be interested to hear where you are sourcing yours from? I’m also in UK and own a 3D printing company. We could make scanning cups and containers, if we could figure out which material to use. Any thoughts?

 

CP – any recommended material here?

 

Paul

[Paul Hiscoe]

Roger I saw your subsequent post where you were considering using PTFE.

 

I just wondered if you have had a chance to try this yet and whether the results are better than with white paper?

Paul

 

[Author]

Posts