Don't Blame Birds for 1918 Flu

[gsgs]

Re: Don't Blame Birds for 1918 Flu

FT is very slow for me currently

niman, can you present the recombination argument ?
What are the parents, what the child.
Is it just only because some part of early sequences appeared in the Canadian swine ?
You often reiterate _that_ (you think) there was recombination,
but no details.

------------------

assume an evolution like this in some segment:

      /-----------------------------swine 1932
    /     
  /     1916
 a------b-----BM 1918
1913     \
          \
           \----------------------------human 1935

then the differences which BM accumulated since its evolution from b
should show up as positions, where both SW and WS should differ
from BM.
But this is not seen, no or only few such positions exist.


I'm creating early consensus/index-sequences for early swine,early human,
1918, and count the mutations...

---------------

OK, I created these indices, http://magictour.free.fr/panflu/i_early
markers in early human (as compared with BM/18): 73+50+58+88+37+54+15+15=390
markers in early swine: 80+88+79+56+47+64+32+27=473
common markers: 10,4,6,7,6,10,5,3 ---corrected:9+3+6+6+4+9+5+3=45

lengths of segments:2280,2274,2151,1701,1497,1410,982,838

473 swine markers in ~14 years,390 human markers in ~17 years,
51 common markers

in the evolution model pictured above, b would be early 1916,a early 1913
assuming no reassortments

I'm getting less reluctant to believe their claim for 1918...
the Jamesburg swine is preserved for 10 years ?


keywords: human1935 swine1932 bm/18 A/swine/Jamesburg/1942(H1N1)


--------------------
graph of avian viruses PB2 differences to early human viruses
http://magictour.free.fr/panflu/PB218A.GIF

H5N1 is much closer to SW/31 than to WS/33 in PB2 and NP

differences in the 8 segments in promille:

QH05-BM18:151,158,138,330,134,174,094,088
QH05-SW31:150,171,147,327,140,173,111,122
QH05-WS33:164,167,142,344,155,200,104,106
BM18-SW31:044,046,051,065,044,051,039,045
BM18-WS33:048,032,037,074,042,061,027,029
SW31-WS33:077,071,069,119,073,089,049,067

 

[pjie2]

Re: Don't Blame Birds for 1918 Flu

Phylogenetic analysis can not detect the recombination that you've described in other posts. If you truly believe that you can, then I realise that the issue is not one where you have your own agenda, but rather you don't understand the methods you're using

Well, yes and no. A phylogenetic analysis can detect longer recombination stretches. If the first half of a segment falls into one phylogenetic group, but the second half falls into another, it's a good indication that there's been a recombination. In terms of detecting smaller recombinant segments... well I guess the same applies, but you have less and less confidence in the results as the segment gets smaller.

In the limiting case where you're talking about single SNPs appearing on a new background and attributing that to recombination rather than de novo mutation, then there is no analysis that can give you that information. It becomes an article of faith. Each camp can look at the other and make arguments from incredulity, but that doesn't actually get you anywhere.

Dr Niman can claim (incredulously) that it's just not realistic to say that the same nucleotide substitution occurs over and over again on different backgrounds. Others can claim (incredulously) that it's vanishingly unlikely to have two recombination events sufficiently close to each other that only one SNP shifts between the strains. Personally I lean towards the latter, but I'm not all that convinced that it matters. Given that oseltamivir resistance can be acquired by a single base change, what does it actually signify whether that change occurs via error-prone replication, via template switching, or via the Noodly Appendage of the Flying Spaghetti Monster himself?

Much more important are the selective pressures that cause that mutation to spread and become fixed in a population once it's occurred - whatever the mutational mechanism. I'm certainly not convinced by Niman's "hitch-hiking" explanation. Hitch-hiking can only occur in the absence of recombination - i.e. a given polymorphism is close enough to a selected variant that it does not recombine and thus gets dragged along for the ride. This is not compatible with a recombination rate that's high enough to shuffle single SNPs. The recent data showing pharmacological levels of oseltamivir in the environment suggest a much more likely reason for the spread of the resistance polymorphism.

 

[gsgs]

Re: Don't Blame Birds for 1918 Flu

> In the limiting case where you're talking about single SNPs
> appearing on a new background and attributing that to
> recombination rather than de novo mutation, then there is no
> analysis that can give you that information

sure there is. It had been done earlier here. (no indication)
Two close SNPs should appear more frequently

this sort of "mutation" requires double-infection which can also
be tested and searched for (reassortments)

reassortments should still be more likely than recombinations

 

[HenryN]

Re: Don't Blame Birds for 1918 Flu

Phylogenetic analysis can not detect the recombination that you've described in other posts. If you truly believe that you can, then I realise that the issue is not one where you have your own agenda, but rather you don't understand the methods you're using

Please. Phylogenetic analysis of each gene segment of 1918 will place the sequences with mammalian sequences and will also place 1918 between classical swine H1N1 and human H1N1 for ALL 8 gene segments.

The recombination within each gene segments is demonstrated by mapping regions of identity, which can be done VERY convincingly by just mapping matches between the 1918 sequence and either WSN/33 (represnting human H1N1) and swine/Iowa/15/1930.

I understand both approaches quite well and the data are VERY clear (no "random mutations' required).

 

[HenryN]

Re: Don't Blame Birds for 1918 Flu

> )

reassortments should still be more likely than recombinations

Nonsense.

 

[HenryN]

Re: Don't Blame Birds for 1918 Flu

Much more important are the selective pressures that cause that mutation to spread and become fixed in a population once it's occurred - whatever the mutational mechanism. I'm certainly not convinced by Niman's "hitch-hiking" explanation. Hitch-hiking can only occur in the absence of recombination - i.e. a given polymorphism is close enough to a selected variant that it does not recombine and thus gets dragged along for the ride. This is not compatible with a recombination rate that's high enough to shuffle single SNPs. The recent data showing pharmacological levels of oseltamivir in the environment suggest a much more likely reason for the spread of the resistance polymorphism.

Reality check. Oseltamivir resistance happens in H3N2 and H1N1 and at multiple positions in N1 or N2 (as was seen in Japan several years ago when it was used at sub-optimal levels in children).

The level of resistance in H3N2 in the past several year is ZERO. The number of resistance changes in H1N1 other than H274Y is ZERO. Attributing the sudden appearance and expansion of H274Y to environmental Tamiflu has ZERO support.

 

[HenryN]

Re: Don't Blame Birds for 1918 Flu

Dr Niman can claim (incredulously) that it's just not realistic to say that the same nucleotide substitution occurs over and over again on different backgrounds. Others can claim (incredulously) that it's vanishingly unlikely to have two recombination events sufficiently close to each other that only one SNP shifts between the strains. Personally I lean towards the latter, but I'm not all that convinced that it matters. Given that oseltamivir resistance can be acquired by a single base change, what does it actually signify whether that change occurs via error-prone replication, via template switching, or via the Noodly Appendage of the Flying Spaghetti Monster himself?

Reality check #2. In seasonal flu, Tamiflu usage is rare (50% of Tamiflu usage worldwide is in Japan). Selection of random errors would put the vast majority of resistance in Japan. In January of 2008 Norway reported resistance in H1N1 (and only H1N1 and only at H274Y) at 67%. Japan had a rate of 3%.
Random mutation rates don't change by country, have no predictive value, and are RARELY selected (except under heavy selection pressure like Tamiflu usage).

Acquistion by recombination is predictable and the prediction is that H274Y will jump to pandemic H1N1 and will follow the same type of jumping seen in seasonal H1N1 because of the large reservoir of H1N1 in seasonal flu.

 

[pjie2]

Re: Don't Blame Birds for 1918 Flu

Reality check. Oseltamivir resistance happens in H3N2 and H1N1 and at multiple positions in N1 or N2 (as was seen in Japan several years ago when it was used at sub-optimal levels in children).

The level of resistance in H3N2 in the past several year is ZERO. The number of resistance changes in H1N1 other than H274Y is ZERO. Attributing the sudden appearance and expansion of H274Y to environmental Tamiflu has ZERO support.

OK, but then how can H274Y hitch-hike under your scenario? If it frequently recombines onto novel backgrounds, it should also frequently recombine off the novel background, unless you allow for direct selection on H274Y itself.

Hitch-hiking or recombination: you can't have both.

 

[pjie2]

Re: Don't Blame Birds for 1918 Flu

> In the limiting case where you're talking about single SNPs
> appearing on a new background and attributing that to
> recombination rather than de novo mutation, then there is no
> analysis that can give you that information

sure there is. It had been done earlier here. (no indication)
Two close SNPs should appear more frequently.

Read the bolded word again. I was specifically talking about the limiting case where only a single SNP is claimed to be "recombining" between strains, and thus you can't use data from other close SNPs to demonstrate a genuine recombination.

 

[gsgs]

Re: Don't Blame Birds for 1918 Flu

recombination won't know, where the SNP is.
niman's theory assumes longer,normal recombinations
which however often only hit one SNP because the strains are so similar

 

[pjie2]

Re: Don't Blame Birds for 1918 Flu

recombination won't know, where the SNP is.
niman's theory assumes longer,normal recombinations
which however often only hit one SNP because the strains are so similar

Yes, if strains are more similar, then the distance between successive SNPs will be larger. What does that have to do with my point? If the recombinant segment only transfers a single SNP, then you cannot tell the difference between recombination and de novo mutation.

 

[gsgs]

Re: Don't Blame Birds for 1918 Flu

if a large amount of SNPs were due to recombination, then
we should see overproportionally many double-SNPs

 

[HenryN]

Re: Don't Blame Birds for 1918 Flu

Yes, if strains are more similar, then the distance between successive SNPs will be larger. What does that have to do with my point? If the recombinant segment only transfers a single SNP, then you cannot tell the difference between recombination and de novo mutation.

The SNP's form patterns. If you are looking at ONE example, you can't say much. When the same polymorphism keeps jumping from one background to another, you can say more. When the jump involves 2 or 3 consecutive polymorphisms, you can say even more. When the acquired polymorphism is from a co-circulating strain, you can also say more. If you look at a get deal of data, the mechanism becomes obvious.

For 1918, which is the topic of this thread, many of the polymorphisms on 1918 are relatively rare, yet they keep showing up in WSN/33 or swine/Iowa/15/1930. Moroever, for 1918 there are doubles and triples in addition to SNPs.

"Random mutation" only survives in the absence of data. If you actually LOOK at the data, recombination becomes abundantly clear and "random mutation" become utter nonsense.

 

[HenryN]

Re: Don't Blame Birds for 1918 Flu

if a large amount of SNPs were due to recombination, then
we should see overproportionally many double-SNPs

No. The VAST majority of recombination is between closely related sequences and the length of the acquired sequence is relative short. The vast majority of the exchanges are SNP's.

 

[HenryN]

Re: Don't Blame Birds for 1918 Flu

Read the bolded word again. I was specifically talking about the limiting case where only a single SNP is claimed to be "recombining" between strains, and thus you can't use data from other close SNPs to demonstrate a genuine recombination.

You can use the frequency of the polymorphism.

 

[super_flu]

Re: Don't Blame Birds for 1918 Flu

pjie2

I love where you've taken this conversation. These are some excellent points that I'm happy to be address

Well, yes and no. A phylogenetic analysis can detect longer recombination stretches. If the first half of a segment falls into one phylogenetic group, but the second half falls into another, it's a good indication that there's been a recombination. In terms of detecting smaller recombinant segments... well I guess the same applies, but you have less and less confidence in the results as the segment gets smaller.

In the limiting case where you're talking about single SNPs appearing on a new background and attributing that to recombination rather than de novo mutation, then there is no analysis that can give you that information. It becomes an article of faith. Each camp can look at the other and make arguments from incredulity, but that doesn't actually get you anywhere.

Your comments bring up two fascinating points.It just so happens I've been re-reading Felsenstein's book, inferring Phylogenies. Excellent reference for anyone that's interested.

The first comment I want to make is that my arguments are not based on my non-belief of a result. I am aware of the assumptions of a phylogenetic analysis and am aware of what can and can not be interpreted from these results. My "incredulity" is of non-data and mis-interpretation of phyogenetic analysis.

You are absolutely correct that conflicting signals in two halves of a sequence may be highlighted by a phylogenetic analysis. However, unless these hypotheses are systematically tested, traditional phylogenetic analysis itself will not detect them (possibly a network analysis may highlight recombination between two lineages). However, in concert with other methods of recombination detection (ie alignement based methods such as bootscan) a phylogenetic analysis is incredibly powerful. This is supported throughout the literature (I can provide specific examples if needed). However, most phylogenetic analyses do not take recombination into account in the models used.

My arguments are made within a phylogenetic framework. This is not a haphazard ordering of data, this is a statistically robust method of analyses. All my arguments come from this framework.

Dr Niman can claim (incredulously) that it's just not realistic to say that the same nucleotide substitution occurs over and over again on different backgrounds. Others can claim (incredulously) that it's vanishingly unlikely to have two recombination events sufficiently close to each other that only one SNP shifts between the strains. Personally I lean towards the latter, but I'm not all that convinced that it matters. Given that oseltamivir resistance can be acquired by a single base change, what does it actually signify whether that change occurs via error-prone replication, via template switching, or via the Noodly Appendage of the Flying Spaghetti Monster himself?

The questions should be focused around the type of data required to detect the recombination Mr Niman describes. Sequencing technologies used currently for genomic sequencing of influenza viruses is absolutely not appropriate to test the hypotheses of template switiching or single nucleotide recombination hypothesized by Niman. Possibly 454 sequencing applied specifically to detect quasi species may provide a statistical measurement of variability at a single nucleotide position. However, the methods used to purify the sample material, amplify genomic material and sequence genetic regions minimize variation in a population from a single individual. It is not just the methodology which can not detect single nucleotide polymorphisms, the data analysed is entirely inappropriate.

Datasets and experiments could be designed to address the recombination that Mr. Niman describes, but so far that data does not exist.

While I do believe you are correct in your assessment of the methodologies, my arguments do not arise from incredulity but rather the lack of data or evidence to test the hypothesis is any statistical framework. If the proper evidence was presented I would believe because as a scientist I believe in evidence.

Much more important are the selective pressures that cause that mutation to spread and become fixed in a population once it's occurred - whatever the mutational mechanism. I'm certainly not convinced by Niman's "hitch-hiking" explanation. Hitch-hiking can only occur in the absence of recombination - i.e. a given polymorphism is close enough to a selected variant that it does not recombine and thus gets dragged along for the ride. This is not compatible with a recombination rate that's high enough to shuffle single SNPs. The recent data showing pharmacological levels of oseltamivir in the environment suggest a much more likely reason for the spread of the resistance polymorphism.

Again, you've asked a fascinating question. How (and why) are mutations selected for (or against) and why do novel mutations become fixed in a population. The most powerful methods currently available to test such questions are based on a sound phylogenetic framework. But, even when shifting to population genetics, the answers regarding frequencies of phenotypes should be based on a phylogenetic framework or the interpretations lose much meaning. - I'm quite tired and will add another post to address this point specifically later.

 

[gsgs]

Re: Don't Blame Birds for 1918 Flu

silly IMO to use phylo-trees to detect recombination.
Write a program and let the computer do it.

 

[HenryN]

Re: Don't Blame Birds for 1918 Flu

pjie2

I love where you've taken this conversation. These are some excellent points that I'm happy to be address



Your comments bring up two fascinating points.It just so happens I've been re-reading Felsenstein's book, inferring Phylogenies. Excellent reference for anyone that's interested.

The first comment I want to make is that my arguments are not based on my non-belief of a result. I am aware of the assumptions of a phylogenetic analysis and am aware of what can and can not be interpreted from these results. My "incredulity" is of non-data and mis-interpretation of phyogenetic analysis.

You are absolutely correct that conflicting signals in two halves of a sequence may be highlighted by a phylogenetic analysis. However, unless these hypotheses are systematically tested, traditional phylogenetic analysis itself will not detect them (possibly a network analysis may highlight recombination between two lineages). However, in concert with other methods of recombination detection (ie alignement based methods such as bootscan) a phylogenetic analysis is incredibly powerful. This is supported throughout the literature (I can provide specific examples if needed). However, most phylogenetic analyses do not take recombination into account in the models used.

My arguments are made within a phylogenetic framework. This is not a haphazard ordering of data, this is a statistically robust method of analyses. All my arguments come from this framework.



The questions should be focused around the type of data required to detect the recombination Mr Niman describes. Sequencing technologies used currently for genomic sequencing of influenza viruses is absolutely not appropriate to test the hypotheses of template switiching or single nucleotide recombination hypothesized by Niman. Possibly 454 sequencing applied specifically to detect quasi species may provide a statistical measurement of variability at a single nucleotide position. However, the methods used to purify the sample material, amplify genomic material and sequence genetic regions minimize variation in a population from a single individual. It is not just the methodology which can not detect single nucleotide polymorphisms, the data analysed is entirely inappropriate.

Datasets and experiments could be designed to address the recombination that Mr. Niman describes, but so far that data does not exist.

While I do believe you are correct in your assessment of the methodologies, my arguments do not arise from incredulity but rather the lack of data or evidence to test the hypothesis is any statistical framework. If the proper evidence was presented I would believe because as a scientist I believe in evidence.




Again, you've asked a fascinating question. How (and why) are mutations selected for (or against) and why do novel mutations become fixed in a population. The most powerful methods currently available to test such questions are based on a sound phylogenetic framework. But, even when shifting to population genetics, the answers regarding frequencies of phenotypes should be based on a phylogenetic framework or the interpretations lose much meaning. - I'm quite tired and will add another post to address this point specifically later.

You are just bouncing around with long irrelevant comments. The 1918 sequence has been published. Some regions have been sequenced in multiple isolates. WSN/33 and swine/Iowa/15/31 have also been sequenced multiple times. The polymorphisms across all 8 gene segments form the same relationship with the two parental sequences. These two sequences alone account for approximately 90% of the polymorphisms in the 1918 strain. Many of these polymorphisms are rare, yet are present almost exclusively in the swine or human sequence, which cam be seen by mapping the matches along each gene.

Phylogentic analysis confirms a human/swine origin for each of the 8 gene segments.

Thus, the phylogeny is used to place 1918 in a mammalian group, which falls been human and swine H1N1, while analysis of the origin of each polymorphism shows that the sequence was generated almost exclusively from human and swine parents which are closely related to WSN/33 and swine/Iowa/15/1930.

 

[gsgs]

Re: Don't Blame Birds for 1918 Flu

name parents,child,segment,region of the recombination.
Or just name a SNP, which you think is created by recombination

 

[HenryN]

Re: Don't Blame Birds for 1918 Flu

name parents,child,segment,region of the recombination.
Or just name a SNP, which you think is created by recombination

1918 is the child and WSN/33 and swine/Iowa/15/1930 are the parents (as has been said MANY times on this thread alone). There are dozens of polymorphisms on each gene segment. Just blast a segment to get the polymorphisms.