The science behind Benghazi

Five Bulgarian nurses and a Palestinian doctor (dubbed the Tripoli six) may be executed soon by the Libyan government for the crime of deliberately infecting over 400 children with HIV. If they did infect the children, this would be a horrendous crime. If they did not infect the children, it's the Libyan government that will be killing innocent people.

The clock is ticking.

Some of you might be wondering (I know I was): How exactly is molecular sequence data being used to solve the crime? Why are scientists and science bloggers claiming that the Tripoli six are innocent?

Let's begin by summarizing what we know.

What do we know about HIV?
HIV evolves at a pretty quick rate. This rate can differ depending on the strain of the virus. This makes it possible for us to do the following:

• 1. Compare the sequences of different isolates. Because HIV changes quickly, we can conclude that if sequences are very similar, then they probably came from the same source.
• 2. Generate possible histories of HIV sequence evolution and use these to calculate the rate of evolutionary change. We can use the estimated rate of change to determine the date when strains diverged.

The most recent paper, published in Nature (1), included work to address both of these points. The best scientific blogging review of first point was by Evolgen. He discussed the phylogeny of different HIV and Hepatitis C isolates and explained why they were likely to have originated from poor hygiene practices at the hospital. See the stories by Revere and Decan Butler for articles that summarize the complete story.

I'm going to add one item from the paper that wasn't presented by others and later on, discuss some evidence from the past, that I think, does a better job of supporting the innocence of the Tripoli six.

Estimating the time of infection
The Nature paper provides the best explanation of the methods used to estimate infection time in the supplementary materials (2). The figure below summarizes their results. In essence, the authors were able to estimate the dates when different viral strains entered the Benghazi patient population. They also determined that the probability of entry after March 1998 (when the Tripoli six came to the hospital) was extremely low.

I like this work, but it involves some very sophisticated math and I wonder how well these kinds of arguments hold up in a courtroom. Will they convince a judge and jury? I hope I'm wrong, but I don't think they'll buy it.

What about the other data?
This might strange, since no one else seems to be writing about it, but I found the most compelling evidence in a 2002 paper by Visco-Comandini, et. al (3). This paper also had a great advantage in that it was easy to get the same data that they used, from the PopSet database, and put it to the test.

The Visco-Comandini study looked HIV sequences from some of the 402 infected children, along with HIV sequences from 19 mothers of infected children, and two infected nurses from the Benghazi hospital.

One of the most important points to remember from the samples in this study is that the mothers had NOT been subjected to any activity that involved drawing blood or sticking them with a needle. Any HIV sequences found in the mothers, then, did NOT originate in the hospital.

I wanted to know: How closely related were the HIV sequences from the mothers, and the nurses, to the HIV sequences from the infected children?

My reasoning was this: If the children were infected with HIV from the hospital, the sequences of the children might be more closely related to other children, and the nurses, than to the mothers.

Visco-Comandini et. al. did this same kind of analysis, in much greater detail and more thoroughly (3), but I wanted to see the results for myself.

I combined the best of these data (most complete sequences) with a reference sequence from Cameroon for the same strain of HIV. Then I opened the file in JalView and generated a multiple sequence alignment with ClustalW. I trimmed off the ends of any incomplete sequences, colored the sequences by the percent identity and generated a tree (using Neighbor Joining).

What do the data say?
The first result that you can see is here:

This image shows a portion of the alignment between these sequences. The most important point to notice is that the top sequence (from Cameroon) looks different from the other sequences. (Purple areas are identical, white shades indicate a base that's only in one sequence.) All of the other sequences (from the Libyan nurses, mothers, and children) are very similar to each other.

It turns out that all of the Libyan sequences are very closely related to each other (> 95% for many of the pairs). This results strongly confirms the findings from others -all of these Libyan HIV strains had a common origin. The mothers, the nurses, and the children were all most likely infected from the same source.

I think these data clear the Tripoli six, since I don't think that they could have infected the mothers or the two infected Benghazi nurses.

These results are from a quick and dirty phylogenetic analysis, but I would really like to know more about the two Benghazi nurses who appeared in the data set and whether or not they had contact with the children.

The tree that I made is shown below. Trees are another way to view and compare sequences. You can see that the sequence from Cameroon, and for one of the children, are quite different from the others. You can also see that the sequences from the mothers, nurses, and children, are very much alike, since they all cluster together.

My fellow ScienceBlogger Mike Dunford, has written a nice description of how you can help and possibly save the lives of the six medical volunteers. Their time is running out.

References:
1. Oliveira, T. et. al. 2006. Molecular epidemiology: HIV-1 and HCV sequences from Libyan outbreak. Nature advance online publication, Dec. 6th. doi:10.1038/444836a2.
2. Benghazi Study Group, Supplementary Materials to Oliviero et al., Nature online, 6 December 2006; cites omitted.
3. Visco-Comandini, et. al. 2002. Monophyletic HIV Type I CRG02-AG in a Nosocomial Outbreak in Benghazi, Libya. AIDS Research and Human Retroviruses 18: 727-732.

technorati tags: Benghazi, Tripoli Six, phylogeny, sequence analysis, HIV