how many organic compounds ?

[Peter Apps]

To help make a point about chemical signalling I need some kind of reasonably authoritative number for how many organic compounds could potentially be in an odour mixture. The requirement that they be in the vapour phase at nanomole concentrations or above limits molecular size to less than about C30, with no multiple polar functional groups (so sugars for example are excluded).

There are all sorts of estimates for the total possible or actual organic structures that do or could exist, but none that I can find that are limited to "small" molecules. Obviously the number is mind-bogglingly huge, which is exactly the point that I want to make !

Since this has to go into a peer reviewed paper I need something that will stand up to scrutiny - an authoritative figure just for hydrocarbons up to C whatever will also make the point.

Can anyone give me a literature reference, or a link to an authoritative website ?

Thanks Peter

[GOM]

Hi Peter,

I used to analyse many fragrances and malodours. Often I would stop identifying after about 100 to 150 components. I could have dived deeper into the chromatogram but it usually wasn't necessary. This paper on the volatiles from a cantaloupe identifies 139 components but doesn't say how many more were unidentified!
http://postharvest.ucdavis.edu/datastor ... 4-1827.pdf

In all the papers that I have read on odours and malodours I don't recall seeing much more than that number identified.
Regards
Ralph

In chemical signalling, I suppose which molecules have an odour (i.e are detected selectively and elicit a response) depends on the species of animal i.e. different species have evolved a sensitivity for particular molecules depending on the question
Can I mate with it?
Can I eat it?
Will it harm me?
That sensitivity will be linked to a response of seek out or flee.

[Peter Apps]

Hi Ralph

The number of compounds identified in any particular complex natural sample is always limited by the technology, and how fussy you need to be about the reliability of the indentification. If you take a good separation on a 30 m capillary, set a low peak threshold and just let the library search run automatically you can get anywhere between 100 and 500 compounds on the list. Add the requirement that the retention indices have to match literature and that drops considerably. constrain it further by needing co-injections of authentic standards and ot drops further still - how far it drops depends on what you can synthesise or buy off the shelf, not on the nature of the sample itself.

In African wild dog scent marks, which is what I am working on, we typically get chromatograms with over a hundred peaks, and we have MS fits with retention matches for just over 100 compounds, with 40 or so confirmed. But on every chromatogram there are a host of "no hits" on library searches, either because the compound is not in the library, or because the peaks are not resolved enough. So the total chemical diversity is way higher than we can measure. Most natural products will give a similar picture - in substances that have really been looked at in detail, like coffe or tobacco smoke, no matter how deep you dig you just keep finding more and more compounds at lower and lower concentrations. From my previous life I recall one flavour peak whose spectrum gave a very pooor fit to something whose literature odour was nothing like what I got on GC-sniffing, when I ran it by heart cutting GC-GC it turned into six compounds.

Conventional wisdom in semiochemistry is that chemical signals of individual identity (or genotype) are necessarily coded by mixtures of chemicals, since there are a very large number of individuals. The counter-argument that I want to make is that there are plenty of chemicals (even just volatile ones) for every individual to have its own single compound. This is part of a larger review on research strategy in mammal semiochemistry.

Peter

[DSP007]

Good morning Peter!
Chemical signalling for the "blind, deaf and dumb" or what?
Mustard oil, capsaicin, allicin, and CN than bad?


Oh( from the second communication) understand the purpose (how realizate) - it's not about the chemical signalling , but of the signaling system (Fairmont) warm-blooded animals.

Immediately come to mind are sex hormones, primarily estradiol.

[Peter Apps]

Hi DSP007

It is indeed about chemical signalling; sending nformation between animals using chemicals, technically known as semiochemistry. "Pheromone" is the term that most non-specialists know best, although most mammal semiochemicals are not strictly pheromones.

Steroid hormones are primarily for signalling within (not between) animals, although steroid metabolites and conjugates in urine are probably what signals reproductive condition of females to males. Being mostly non-volatile these are detected by the vomeronasal organ (Jacobsen's organ) rather than by the main olfactory system.

Peter

[Consumer Products Guy]

The fragrances used in our consumer products typically contain 50-100 added components. However, many of those components are not pure, so I'd estimate 100-150 readily-observable components, but more if one separates better and looks for trace components. Many of the ingredients are natural extracts like essential oils, big mixtures.

[Peter Apps]

Another thing to keep in mind is that the LOD of a dog's nose gets down to 10 to the minus 18 molar, so the peaks that anyone sees on a chromatogram of a natural product are only the concentrated tip of the chemical diversity iceberg.

There are formulae for working out the number of isomers of fairly simple structures like alkanes (just over 4 billion for C30H62) http://www-jmg.ch.cam.ac.uk/data/isomercount/, so I suppose that I could add all those up for the various alkane lengths and use that as a minimum figure.

Peter

[tom jupille]

Peter, you're waay ahead of me (as usual ). I was going to suggest that if all you need is an order-of-magnitude estimate, just start with hydrocarbons and assume that each carbon is bonded to three other atoms (that fudges things down a bit for double bonds, aromatics, and hetero atoms). That gives 3^n where n is the number of atoms you want to limit yourself to. If you use n=15 as a SWAG ("Scientific Wild-Assed Guess) for a lower bound , that works out to about 10^7.

But then, to quote Mark Twain (writing about the decreasing length of the Mississippi River):
There is something fascinating about science. One gets such wholesale returns of conjecture out of such a trifling investment of facts.

No matter what number you come up with, it's going to be a *lot* higher than the peak capacity of a GC column!

[GOM]

Peter,

All of my studies were carried out on humans. Out of interest, have you been able to carry out any studies on identical twin dogs?

[GOM]

Peter,

Just a thought, but what is the size of the current NIST library? That could be a starting point even allowing for duplications.

Regards,

Ralph

[Peter Apps]

Hi Ralph

As far as I know nobody has ever looked at identical twin dogs - I am not even sure that dogs even have identical twins ! There was the classic study that showed that dogs cannot discriminate identical twin humans eating the same diet.

Inbred strain lab mice are as identical as identical twins, and can be trained to discriminate between one anothers odours. This is probably due to odours from the major histocompatability complex - the genes that programme the immune system. In wild mice individual recognition is based on major urinary proteins, which are signallers in their own right and carriers of volatile semiochemicals that have all sorts of behavioural and physiological effects.

I thought about the NIST or Wiley MS libraries, but some pedant will say that a large number of the compounds are man-made (which is true).

Hi Tom,

The likeliehood that semiochemicals are obscured by other peaks on the chromatogram will get a mention. Cal Giddings' papers on the (in)capacity of capillary columns to separate what could be in a sample should be required reading for every chromatographer working with complex samples. If I recall the biggest alkane hydrocarbon where all the isomers can be separated on a single column is hexane.

I plan to use the astronomical numbers of chemical compounds to specifically refute the argument that signals of individual identity have to be mixtures. If there are more chemicals than there are individuals (which is true by anyone's estimate, even with sensible constraints on structures based on metabolic pathways) then in principal every individual could have its own single-compound label. This is not to say that they do have single compound labels, just that it is in principal possible, and that we should not use methods that stop us from finding them.

Peter

[lmh]

Peter, can I just chip in?

(1) When I first read your question, my immediate thought was "great, that's the sort of question Peter Apps will be able to answer for this chap"...

(2) I'd be a bit wary, from a psychological point of view, of what you're trying to do. If I were the peer reviewer, my thoughts would be along the following lines: "Fine, it's theoretically possible to have a situation where every individual is identified by a single chemical totally unique and found in no other, but how is this achieved biochemically? Are you expecting me to believe that every individual has made up some chemical pathway absent from every other? That every individual has a unique synthetic pathway. This sounds like a very un-Occam's-razor approach, an unnecessarily complicated hypothesis. Nevertheless, you're right to question whether we should set ourselves up in a way that prevents us from noticing if it's true. But is this a case of an analytical expert arguing for bigger and better equipment and more money for developing methods that are more sophisticated than we really need (arguments like "Biology is complex, so give me more money please" ring alarm bells for me)? Surely if every individual has their own chemical, we'll notice pretty soon that we're seeing novel peaks, even if we can't actually separate every novel peak. We don't need a method that will separate 4 billion chemicals, we just need a method that has enough blank space between its peaks to provide a 'trap' in which to catch a new chemical, if one exists."

In effect, I'd regard this very large number as sort-of-interesting but not really changing how I felt about a paper. I'd be more alarmed by the dog's-nose-LOD issue.

(3) Practically I don't think you can answer your question. You can estimate how many chemicals are theoretically possible, as you have tried, but you can't estimate how good we need to make our method if we wish to see all chemicals that really exist, because we only know they really exist if we have a method that can see them. It's currently a circular argument.

[Peter Apps]

Hi lmh

You make some very good points (with the possible exception of the first )

Trust me, I am likely to have worse problems with the reviewers than the biochemistry of chemical diversity ! The argument that I am making is the opposite of the current orthodoxy, so I am expecting a very exacting time. It is the constraint on biochemical pathways that makes me prefer some kind of sensible estimate for how many (natural) volatile organic compounds there really are, rather than how many there could be. Actually the exercise has already been done for insect pheromones (John A. Byers 2005. Chemical constraints on the evolution of olfactory communication channels of moths. Journal of Theoretical Biology 235: 199-206.) - there are tens of thousands of possible isomers of a small subset of real moth pheromones. An additional constraint is whether a nose (or vomeronasal organ) can discriminate between different compounds. Luca Turin argues that no two compounds smell exactly the same (even some optical isomers smell different) but that has about as much empirical content in it as the estimate of 4 billion isomers for a C30 alkane. Since each animal only interacts with a tiny fraction of the whole population of its species (at least those interactions where individual identity makes a difference) individual recognition could be organinsed with only a couple of thousand single compound name tags - and even then there would be plenty of spare capacity. To argue that though I need a plausible and authoritative number for the size of animal social networks - one extreme is the recognition (by odour) of mexican free-tailed bat mothers of their own pups who roost together in creche groups of 4000. Isn't nature wonderful !

In this paper I am not arguing for increased analytical power (that will be in another one ), I am arguing against the current popularity of multivariate number crunching to find patterns of peak area ratios on raw chromatograms. In principal this could work, but nearly everyone who does it first throws out the peaks that do not occur in all or most of the chromatograms, which to my way of thinking is the same as ignoring eye and hair colour in facial recognition because not everyone is a blue-eyed blonde. IF the signals are single compounds or simple mixtures where the compounds are either present or absent, then the peaks for the signal compounds do not even make it as far as the statistical analysis.

Peter

[Consumer Products Guy]

Peter,

All of my studies were carried out on humans. Out of interest, have you been able to carry out any studies on identical twin dogs?

Fraternal or identical twin dogs?

[GOM]

Hi CPG,

I meant identical. Like Peter I had an hypothesis which went against current thinking. Ultimately the only test of an hypothesis is by experiment. Unlike Peter I had the luxury of being able to put a human on the end of a GC column to sniff ( other workers in the insect field can attach an antenna to the outlet.) It took 3 years of hard work to confirm our hypothesis. I had to pull out every trick that I had ever learnt in of 30 years of specialising in GC and invent some new ones but we did it - it really did feel good when we proved it!

Cheers

Ralph