First, a univariate analysis was carried out, which showed that the number of changes in the P1 and VP4 proteins did not correlate to in-vitro cross-protection, whereas a link was evident for the three surface-exposed proteins (VP1-3), with Rigosertib VP3 showing the strongest association (P < 0.001). A subsequent multivariate analysis to evaluate the three different VP regions and their interactions did not identify any significant interactions. Changes in VP3 and VP2 showed a significant (negative) effect on the probability
of protection; the higher the number of changes the lower the probability of protection (Supplementary Table 2). The absence of a relationship between predicted protection of vaccines and changes in capsid aa of field viruses observed in our analysis is in keeping with other evidence that neutralisation is governed by key (mutant-) capsid aa residues, and probably by residue interactions, rather than overall residue changes [10]. However, the observation of a relationship between predicted protection and the substitution of aa in VP3 is interesting. Assessing the contribution of specific substitutions to predicted cross-protection requires more advanced analytical approaches and manipulation of selected
aa residues using reverse genetics approaches. The multivariate analysis also allowed a comparison of the predicted Perifosine cell line level of cross-protection provided by each of the commercial and candidate vaccine strains used in this study. A-EA-2007, A-EA-1984 and A-EA-1981 exhibited significantly higher expected protection with A-EA-2007 exhibiting the highest odds value (Table 3). A-ETH-06-2000 was not significantly different from A-ERI-1998, while A-KEN-05-1980 was significantly less protective than A-ERI-1998. The vaccines (A-ETH-06-2000 and A-KEN-05-1980)
showing the lowest in-vitro cross-protection based on r1-values ( Fig. 1) also showed the lowest odd values ( Table 3). In conclusion, two isothipendyl topotypes (African and Asian) of the type A viruses were detected in East Africa; of the native African topotype three genotypes are currently circulating in the region. We have recommended different vaccines for the different genotypes based on their serological cross-reactivity and genetic relationship. A-EA-2007 has broader cross-reactivity and is also a recent isolate; therefore, is recommended as a potential vaccine strain candidate to be used in FMD control programs in East Africa, subject to good growth and stability characteristics and in vivo evaluation in the target host. We would like to thank WRL-FMD at Pirbright for providing the viruses for this study and Dr Gelagay Ayelet, National Veterinary Institute, Ethiopia for sharing vaccine sera. The authors thank Dr J. Gonzales for help with GLM analysis. This work was financially supported by BBSRC, DFID (Grant nos. BB/H009175/1 and BB/F009186/1).