Skip to main content

Table 5 Computational times

From: Multi-objective optimization for RNA design with multiple target secondary structures

   MODENA   
Dataset l (nt) pop. size #threads t MODENA (s) t Frnakenstein (s) t RNAdesign (s)
two targets 149 100 4 446a 13,988a 2,924c
three targets 100 100 4 339a 1,924b 1,249c
four targets 100 100 4 385a 2,848b 1,653c
RNA deviced 95 100 4 2,428a - -
RNA devicee 95 200 6 15,460b - -
PK60 60 30 3 1,997b - -
PK80 80 30 3 7,780b - -
LE80 61 30 3 2,140b - -
  1. An l indicate the mean length for each dataset. The ‘pop. size’, ‘#threads’, and t MODENA columns indicate GA population size, the number of OpenMP threads used to parallelize a design, and the mean computational time of MODENA for each dataset, respectively. For example, from the ‘two targets’ row of this table, we can see that one MODENA run of the two-target design with a population size of 100 and four OpenMP threads took 446 seconds on average. PK60, PK80, and LE80 are the two-target datasets with pseudoknot. The computational times for “frnakenstein.py -s 100” and RNAdesign with option “-n 500 –thin 200 -b 100 –scale 1” are shown in t Frnakenstein and t RNAdesign columns, respectively
  2. aComputational times measured on a PC with Intel Xeon E5-2603 (1.80 GHz) and 16 GB of memory (CentOS 5.9)
  3. bThe times measured on a PC with Intel Xeon E5-2665 (2.40 GHz) and 132 GB of memory (CentOS 6.4)
  4. cComputational times measured on a PC with Intel Core i3-2100 (3.10 GHz) and 3.6 GB of memory (Fedora 18, which was installed in order to execute RNAdesign)
  5. dand e indicate RNA device designs which take and does not take energy barrier height into account, respectively
  6. euses the maximum number of GA generations of 400