Talk:AIDS/Archive 11

From Wikipedia, the free encyclopedia

[edit] Disputed dating method of HIV-1 genetic ancestry dating

Lestatdelc June 28, 2005 22:25 (UTC) As noted in www.aidsorigins.com there are serious flaws dating method of HIV-1 genetic ancestry dating method citing the the Science 2000 Jun 9 article and the results are very much in dispute:

The article you quoted extensively (titled "Robin Weiss, professor of virology, doctor of spin") from below comes from Aidsorigin.com. Aidsorigin.com is site devoted to promoting the Oral Polio Vaccine (OPV) hypothesis of the entry of HIV into the human population. The results of phylogenetic dating of the beginings of HIV are only "very much in dispute" among proponents of the OPV hypothesis. In the mainstream of HIV science, it is generally accepted.
The OPV hypothesis alleges that the origins of HIV in the human population can be tranced oral polio vaccine that was contaminated with SIV and given out in the Congo during 1957-1960. Evidence that places the origin of HIV well before the mid-to-late 1950s is disconfirming to the OPV hypothesis. Thus, the peoponents of that theory have to find reason to reject any such evidence.
I would also like to point out that the Science article I cited, and that "Lestatdelc" has taken issue with ("Timing the Ancestor of the HIV-1 Pandemic Strains"), is not the only paper to come to a similar conclusion about the date of HIV's origin. Take a look at:
Carl Henderson 29 June 2005 04:56 (UTC)
Reference for the original study cited:
To summarize, I've cited four different peer-reviewed studies that came to very similar ranges of dates for the origins of HIV-1. I have not added "mutiple reference to the SAME disputed study."
MRCA HIV-1 Group O 1920 (1890-1940)--Korber, et al. (2000)
MRCA HIV-1 Group M 1931 (1915-1941)-Lemay, et al. (2004)
MRCA HIV-1 Group M "before the 1920s–1930s"--Salemi, et al. (2001)
MRCA HIV-1 Group M "late 1930s"--Sharp, et al.(2000)
(MRCA = Most Recent Common Ancestor)
Please note that these are not the only studies coming to similar conclusions. These are just the only ones I was able to find PDFs of on the web (if anyone wants a copy of them to review, please email me at jch@carlhenderson.net). Each of those studies above references others that the more dedicated may wish to search out in the cold, scary offline world of dusty library stacks. I think I have good reason for inserting the line concerning the dating of HIV-1 origins, and good reason for reverting an attempt to delete it.
Carl Henderson 30 June 2005 21:12 (UTC)

From the www.aidsorigins.com article:

Making misleading claims about phylogenetic analysis. Professor Weiss claims that "the most telling answers" about the provenance and timing of the HIVs come from phylogenetic analysis. He argues that phylogenetic dating confounds the OPV theory, since it indicates that the Most Recent Common Ancestor (MRCA) of this group existed "around 1931, with 95% confidence limits of 15 years". Weiss is correct with regard to the provenance of the HIVs, since phylogenetic analysis provides important evidence about which SIVs are their closest non-human relatives. However, the use of phylogenetic analysis to date the epidemic is highly controversial, as Weiss is well aware. Significantly, he makes no reference to this.
This omission is doubly surprising, given that (a) his fellow-organiser from the Royal Society meeting, Simon Wain-Hobson, has revealed major shortcomings in this aspect of phylogenetic theory [see below], and (b) Professor Weiss heard the presentation made at Lincei by Mikkel Schierup, which focused on the shortcomings of phylogenetic dating.
In his conclusion, Weiss states that Schierup's paper [Schierup and Forsberg, 2003] merely suggests that "the confidence limits might widen a little" on that date of 1931 plus or minus 10 to 15 years. However, this is a misrepresentation of what Schierup and Forsberg write in their important, indeed ground-breaking, paper.
They report that there is substantial evidence of recombination having occurred both between and within the subtypes of HIV-1, and they find evidence of extensive recombination even in datasets such as that used by Bette Korber in her famous 2000 paper, "Timing the Ancestor of the HIV-1 Pandemic Strains" [Korber et al., 2000], which the authors claimed to have "cleaned" of potential recombinant sequences. Schierup and Forsberg demonstrate that ignoring this recombination would lead to placing the MRCA too far back in time.
This conclusion has since been greatly reinforced by the work of Andreas Meyerhans and Simon Wain-Hobson, who report evidence of "massive" recombination in HIV-1 [Wain-Hobson et al., 2003]. These latter authors calculate that the rate of recombination in HIV-1 is some ten times greater than the rate of "point-substitution" (ie normal mutation). Ignoring recombination, they estimate, would lead to overestimating the age of the MRCA by as much as 45%. (This should not be taken as indicating an upper limit; it may well be that the phylogenetic dating approach would involve considerably greater errors than this.) However, even correcting for a 45% error would bring Korber's mooted MRCA up to about 1950 (range: approximately 1940 to 1957), and Paul Sharp's putative 1940 MRCA up to about 1957 (range: approximately 1948 to 1963). In other words, the OPV trials of the 1950s would now fall within the time estimates.
However, Schierup and Forsberg go much further than that, for they call the whole basis of phylogenetic dating into question. They write that the very high rate of recombination between subtypes "is important, because it indicates that recombination may have been prevalent also before the diversification into subtypes, ie early on in the evolution of the present diversity." Later, in their conclusions, they write that: "If recombination has occurred in the viral population originating from the MRCA [ie early in the evolution of HIV-1], it is not valid to use a phylogenetic method to obtain the time estimate, and our results suggest that doing so would give a certain overconfidence in the previous estimate of 1931 +/- 10 years". [My bracketed insert and italics.] What Schierup and Forsberg are actually stating here is their belief that recombination probably did also occur early in the history of HIV-1 (recombination that would neither be detectable, nor removable by the "cleaning" methods advocated by such as Korber and Sharp). They are further stating that such early recombination would mean that phylogenetic dating theory is inherently flawed, and should not be used to date HIV-1.
Yet all Professor Weiss hears is that polite, throw-away line about previous studies demonstrating a "certain over-confidence". Weiss was meant to provide a balanced summing-up of the papers presented at the conference, but in fact he has minimised the import of Schierup and Forsberg's work. It is as if he has only read the last two paragraphs of their paper, and has lifted out the one phrase that fits with his preconceptions.
The simple truth is that phylogenetic dating is now revealed as an inappropriate technique for dating HIV-1, as is acknowledged by a growing number of scientists, both in print (Schierup and Forsberg, Wain-Hobson, Meyerhans, Jon Cohen of Science [Cohen, 2002]), and verbally (as at least half a dozen other statisticians and geneticists have done in conversation with me). Weiss's claim that "the confidence limits might widen a little" is disingenuous, and appears to be an attempt to retain some credibility for phylogenetic dating. The only conclusion that can reasonably be drawn from the work of Schierup and others is that phylogenetic dating theory is innately flawed, and therefore unsafe, when applied to HIV-1, and should therefore not be used as a basis for estimating the age of AIDS viruses in humans.
One final point. Weiss states that what convinces him of the accuracy of Korber's start date of 1931 is that this scenario successfully predicts two "known dates of sampling": the 1959 Leopoldville sequence, and the 1986 date for the emergence of subtype E in Thailand. In reality, both claims are questionable. Korber's actual prediction for the date of Leopoldville sequence is 1957, with 95% confidence intervals [95% CI] of 1934-1962. Yet, as I have explained in pages 173-176 of Dephlogistication [Hooper, 2003], there is considerable doubt about the dating of the Leopoldville sample; a wide range of evidence (including testimony from the three main authors of the genetics papers about these samples &emdash; Motulsky, Vandepitte and Fraser) indicates that this sample was certainly obtained later than the declared date of "early 1959", and it may have been obtained as late as 1963 (which would fall outside Korber's confidence intervals).
Again, Korber's best prediction for the Thai subtype E founder sequence (using the full set of Group M sequences) is actually 1984 [95% CI: 1980-1986]. Yet, on the basis of the discussion in her own paper, 1987 would appear to be a more plausible date for the origin of the Thai subtype E epidemic than 1986 &emdash; and again, 1987 would fall outside her confidence intervals.
Korber encountered similar problems with her estimates of the putative date of the first American subtype B sequence (though Weiss omits to mention this). She has four different guesses at this one, but the best guess, based on the envelope gene, places the first US subtype B sequence in 1967 (95% CI: 1960-1971). As several observers, most notably Gerry Myers, have pointed out, this seems far too early when the first confirmed American HIV-1 isolate is from 1977, and the first proven case of American AIDS is from 1979.
In all three instances, therefore, it seems that the real dates of the sequences in question are more recent that Korber's estimates, falling either at the upper limit of her 95% confidence intervals, or outside them. Leaving aside the question of whether phylogenetic dating works, this suggests that her estimates for these dates (including her famous 1931 estimate) are set considerably too far back in time.
Coming back to Robin Weiss, it appears that he has not taken the trouble to recheck the actual dates and confidence intervals in Korber's paper, for the details which convince him that her 1931 estimate is correct, if examined closely, actually do the exact opposite. In reality, they consistently suggest that her 1931 estimate is too early.