figure 5. two favored mutational escape paths of hIV from the therapy with the
Rt inhibitor zidovudine.
tAm1 Path
M
41
l
D
tAm2 Path
gives a normalized value reflecting the
significance of the resistance value.
Column 4 lists mutations found in the
input sequence, red if they strengthen
the resistance of the virus and green if
they weaken it. The data in the figure
points to strong resistance against
many inhibitors of RT and therapy options targeting PR. The Geno2pheno
server is the basis for supporting treatment decisions in about two-thirds of
HIV-infected patients treated in Germany. 13 This means at least 12,000
decisions for treatment selection per
year in Germany involve geno2pheno
or its findings.
Chasing the virus. This analysis
treats each drug separately. Given the
output in Figure 4, the physician as-sesses the resistance level of the virus against each individual drug and
manually composes the combination
drug therapy that is (hopefully) effective against the present virus. We also
look into the future of the virus. Presented with a given combination drug
therapy, how will it react? What are its
mutational escape paths and how long
will the therapy stay effective? The virus does not just randomly introduce
mutations. Rather, it follows more-or-less established mutational escape
paths; Figure 5 outlines two favored
paths from a therapy with the single
AIDS drug zidovudine (ZDV, AZT). (The
notation is analogous to that of Figure
3.) We denote with K70R the mutation
of K to R in position 70 (of RT). Hence,
one escape path is K70R followed by
K219E/Q.
The biological reasons for the vi-
rus following these paths are not well
understood. But the paths show up
in a clinical HIV-resistance database.
Finding them is simple if we have lon-
gitudinal data. The data comprises se-
quences of viral genotypes and clinical
parameters from the same patient over
long periods of time. However, such
data is difficult to come by. Our data-
bases are dominated by cross-sectional
data involving only a few or single data
points for each patient. Nevertheless,
we are still able to identify favored es-
cape paths from cross-sectional data,
as in Figure 5.
figure 6. two mutagenetic trees indicating the escape paths in figure 5; “wild type”
indicates the reference virus.
0.22
wild type
41l 67n 69Dn 70r 210W 215FY 219EQ
+0.78
wild type
tAm2 70r
tAm1 215FY
219EQ
41l
67n
210W
69Dn
MArch 2010 | vOl. 53 | nO. 3 | CommunICAtIonS of the ACm 71
