Until the early 1990s, incidences of Mycoplasma pneumoniae (MP) infection showed three to five year epidemic cycles in multiple countries, however, the mechanism for the MP epidemic cycle has not been understood. Here, we investigate the determinant of periodicity in MP incidence by employing a mathematical model describing MP transmission dynamics. Three candidates for the determinant of periodicity were evaluated: school-term forcing, minor variance in the duration of immunity, and epidemiological interference between MP serotypes. We find that minor variation in the duration of immunity at the population level must be considered essential for the MP epidemic cycle because the MP cyclic incidence pattern did not replicate without it. Minor variation, in this case, is a less dispersed distribution for the duration of immunity than an exponential distribution. Various lengths of epidemic cycles, including cycles typically found in nature, e.g. three to five year cycles, were also observed when there was minor variance in the duration of immunity. The cyclic incidence pattern is robust even if there is epidemiological interference due to cross-immune protection, which is observed in the epidemiological data as negative correlation between epidemics per MP serotype.

The prediction of the lineage dynamics of influenza B viruses for the next season is one of the largest obstacles for constructing an appropriate influenza trivalent vaccine. Seasonal fluctuation of transmissibility and epidemiological interference between the two major influenza B lineages make the lineage dynamics complicated. Here we construct a parsimonious model describing the lineage dynamics while taking into account seasonal fluctuation of transmissibility and epidemiological interference. Using this model we estimated the epidemiological and evolutional parameters with the time-series data of the lineage specific isolates in Japan from the 2010–2011 season to the 2014–2015 season. The basic reproduction number is similar between Victoria and Yamagata, with a minimum value during one year as 0.82 (95% highest posterior density (HPD): 0.77–0.87) for the Yamagata and 0.83 (95% HPD: 0.74–0.92) for Victoria, the amplitude of seasonal variation of the basic reproduction number is 0.77 (95% HPD:0.66–0.87) for Yamagata and 1.05 (95% HPD: 0.89–1.02) for Victoria. The duration for which the acquired immunity is effective against infection by the Yamagata lineage is shorter than the acquired immunity for Victoria, 424.1days (95% HPD:317.4–561.5days). The reduction rate of susceptibility due to immune cross-reaction is 0.51 (95% HPD: 0.084–0.92) for the immunity obtained from the infection with Yamagata against the infection with Victoria and 0.62 (95% HPD: 0.42–0.80) for the immunity obtained from the infection with Victoria against the infection with Yamagata. Using estimated parameters, we predicted the dominant lineage in 2015–2016 season. The accuracy of this prediction is 68.8% if the emergence timings of the two lineages are known and 61.4% if the emergence timings are unknown. Estimated seasonal variation of the lineage specific reproduction number can narrow down the range of emergence timing, with an accuracy of 64.6% if the emergence times are assumed to be the time at which the estimated reproduction number exceeds one.

To estimate and predict the transmission dynamics of respiratory viruses, the estimation of the basic reproduction number, R0, is essential. Recently, approximate Bayesian computation methods have been used as likelihood free methods to estimate epidemiological model parameters, particularly R0. In this paper, we explore various machine learning approaches, the multi-layer perceptron, convolutional neural network, and long-short term memory, to learn and estimate the parameters. Further, we compare the accuracy of the estimates and time requirements for machine learning and the approximate Bayesian computation methods on both simulated and real-world epidemiological data from outbreaks of influenza A(H1N1)pdm09, mumps, and measles. We find that the machine learning approaches can be verified and tested faster than the approximate Bayesian computation method, but that the approximate Bayesian computation method is more robust across different datasets.

BackgroundHerpes Simplex Virus Type 2 (HSV-2) is one of the most common sexually transmitted diseases. Although there is still no licensed vaccine for HSV-2, a theoretical investigation of the potential effects of a vaccine is considered important and has recently been conducted by several researchers. Although compartmental mathematical models were considered for each special case in the previous studies, as yet there are few global stability results.ResultsIn this paper, we formulate a multi-group SVIRI epidemic model for HSV-2, which enables us to consider the effects of vaccination, of waning vaccine immunity, and of infection relapse. Since the number of groups is arbitrary, our model can be applied to various structures such as risk, sex, and age group structures. For our model, we define the basic reproduction number ℜ0 and prove that if ℜ0≤1, then the disease-free equilibrium is globally asymptotically stable, whereas if ℜ0>1, then the endemic equilibrium is so. Based on this global stability result, we estimate ℜ0 for HSV-2 by applying our model to the risk group structure and using US data from 2001 to 2014. Through sensitivity analysis, we find that ℜ0 is approximately in the range of 2-3. Moreover, using the estimated parameters, we discuss the optimal vaccination strategy for the eradication of HSV-2.ConclusionsThrough discussion of the optimal vaccination strategy, we come to the following conclusions. (1) Improving vaccine efficacy is more effective than increasing the number of vaccines. (2) Although the transmission risk in female individuals is higher than that in male individuals, distributing the available vaccines almost equally between female and male individuals is more effective than concentrating them within the female population.

Human influenza pandemics have historically been caused by reassortant influenza A viruses using genes from human and avian viruses. This genetic reassortment between human and avian viruses has been known to occur in swine during viral circulation, as swine are capable of circulating both avian and human viruses. Therefore, avian-to-swine transmission of viruses plays an important role in the emergence of new pandemic strains. The amino acids at several positions on PB2, PB1, and PA are known to determine the host range of influenza A viruses. In this paper, we track viral transmission between avian and swine to investigate the evolution on polymerase genes associated with their hosts. We traced viral transmissions between avian and swine hosts by using nucleotide sequences of avian viruses and swine viruses registered in the NCBI GenBank. Using BLAST and the reciprocal best hits technique, we found 32, 33, and 30 pairs of avian and swine nucleotide sequences that may be associated with avian-to-swine transmissions for PB2, PB1, and PA genes, respectively. Then, we examined the amino acid substitutions involved in these sporadic transmissions. On average, avian-to-swine transmission pairs had 5.47, 3.73, and 5.13 amino acid substitutions on PB2, PB1, and PA, respectively. However, amino acid substitutions were distributed over the positions, and few positions showed common substitutions in the multiple transmission events. Statistical tests on the number of repeated amino acid substitutions suggested that no specific positions on PB2 and PA may be required for avian viruses to infect swine. We also found that avian viruses that transmitted to swine tend to process I478V substitutions on PB2 before interspecies transmission events. Furthermore, most mutations occurred after the interspecies transmissions, possibly due to selective viral adaptation to swine.