Developing and refining methods of analysing malaria genetic data obtained from infected human blood samples.

The spread of drug resistant malaria is a potent threat to human health in much of the developing world. Resistance can be tracked and quantified provided we know the mutations responsible for resistance (true in many cases) and these data form the basis for many key genetic surveillance programmes aimed at ensuring the continued provision of effective drugs. The requisite malaria genetic data can, in principle, be obtained from analysing infected human blood samples but people are often simultaneously infected by several malaria clones and this makes it impossible to clearly identify the genetic composition of any single clone. We propose a program of research to develop and refine suitable methodologies that can overcome these problems of data interpretation allowing good quality genetic data on the frequency of drug resistance to inform public health policy. These methods of analysis also have more general application in understanding the genetic epidemiology of malaria transmission.

The easiest method of obtaining malaria genetic data from natural transmission settings is by genotyping infected human blood samples. There are two major methodological issues in analysing these data. Firstly, we can only identify alleles as being present or absent in the blood sample so that, for example, if a blood sample contains 5 malaria clones and both wildtype and mutant alleles are detected at a locus, it is impossible to discern if that human infection contains 1,2,3 or 4 mutant clones; this precludes estimating population allele frequencies by simple counting. The second problem is that we may detect several alleles at different genetic positions in the same blood sample but we cannot always recover the individual malaria multi-locus haplotypes. These issues become extremely important when trying to assess the level of drug resistance in the malaria populations but have more general application in understanding the genetic epidemiology of malaria transmission. We have developed maximum likelihood (ML) methods to infer allele and haplotype frequencies from infected blood samples but these need to be updated to reflect different assumptions about malaria genetic epidemiology (for example, the possibility of co-transmission of genetically-related clones), to reflect differences in data structure (for example, the number of clones in a human blood sample, the MOI, may not be reported), and to reflect differing assay sensitivities and the presence of genotyping errors in the field datasets. We will address these issues by extending the ML approach and developing a parallel Bayesian approach.

The continued use of chloroquine as a monotherapy in the face of widespread resistance was arguably the greatest malaria scandal in recent years and undoubtedly cost a huge number of lives [Trape, J. F. et al., 1998, "Impact of chloroquine resistance on malaria mortality." Comptes Rendus De L Academie Des Sciences; Trape, J., 2001, "The public health impact of chloroquine resistance in Africa." Am. J. Tropical Med. Hyg.]

The situation only improved after a group of leading researchers published a controversial letter in the Lancet explicitly accusing the World Health Organisation and World Bank of "Medical Malpractice" in colluding in the continuing deployment of this drug (Attaran et al, 2004, "WHO, the Global Fund, and medical malpractice in malaria treatment").The legacy of this bitter controversy has been a keen awareness of the need for the continued provision of effective antimalarial drugs and, in particular, that the same mistakes of continuing to use ineffective drugs must not be repeated if and when resistance arises to the current generation of antimalarial drugs, the artemisinin combination therapies (ACTs). The malaria community has subsequently achieved a consensus based around the need to establish formal processes linking the collection and analysis of field data on drug resistance (largely carried out by academic teams) to feed into policy decisions. Our work forms an integral part of this process through developing more sensitive and accurate early warning systems based around surveillance of genetic data.

A secondary objective will be to improve the quality of surveillance data. There is little guidance on how best to design and implement a malaria genetic surveillance survey and, as far as we know, no guidance based on statistical modelling. We need not be passive users of field datasets and intend to simulate datasets to investigate how their accuracy depends on, for example, sample size, provision of MOI data, and laboratory estimates of assay sensitivity and error rates. This is unlikely to be as methodologically challenging as the factors enumerated above, but we suspect it will be one of the most useful and widely used outputs of the project.

The main beneficiaries will therefore be people living in resource-poor areas who contract malaria (estimated at 300 to 600 million clinical cases per year) and who need access to fully effective drugs. Our research will help inform and drive the decision making process underlying drug deployment policies.