Table 1. Designs for viral metagenome experiments.

S ()		Eq. 3 Reads	Eq. 4 Reads	Eq. 5 Reads	(5,50,95)% minimax
100	4	62402	62166	67109	3.38, 3.68, 3.94
200	4	128399	135996	142673	3.32, 3.60, 3.80
400	4	263645	303155	310081	3.24, 3.49, 3.72
100	5	88636	89303	96992	4.43, 4.85, 5.15
200	5	181745	196402	206985	4.28, 4.70, 5.04
400	5	371999	438059	449314	4.23, 4.56, 4.78
100	6	113767	115738	126271	5.63, 6.16, 6.50
200	6	232749	255203	269879	5.39, 5.92, 6.27
400	6	475413	569204	585113	5.17, 5.73, 6.04

Table 1 provides the numbers of reads of size determined to give 95% probability of assembling contigs of at least size in viral ( = 200000,  = Uniform(50000,350000)) metagenomics problems as a function of the number of species or in the pool. Calculations are provided for models using fixed pool and equal genome sizes and abundances (Eq. 3), fixed pool sizes with distributed genome sizes and abundances (Eq. 4) and stochastic pool sizes with distributed genome sizes and abundances (Eq. 5). (5, 50, 95)% minimax contig size quantiles from simulated assemblies of species with uniformly distributed genome sizes and Pareto distributed abundances using stochastic pool size/distributed genome size and abundance experimental designs are provided for verification. Larger numbers of reads are required to obtain a given level of performance as pool sizes increase, the required performance level increases, if an assumption of equal genome sizes and abundances is replaced with one of distributed genome sizes/abundances with equivalent mean genome sizes, or if a fixed pool size is replaced with a stochastic pool. Consistent with previous observations, minimax contig size quantiles are slightly (less than one read length) lower than planned.