About the design
Figure 1. Schematic view of the leg DNA sequence
The legs of our molecular robot have certain versatility, i.e. being the same leg type or different types. In our first idea about the triangular prism robot we thought it to have three different kinds of legs for performing its rolling motion. However, our simulation program predicted the robot to have a more efficient movement by just having one type of legs . In this section we would like to describe our first design where the robot utilizes three different legs, substrates and field sequences (Figure 1) for its motion mechanism.
For robustness, the system is designed to have two legs of the same type in each of the three edges of the prism shape robot (Figure 2).
The DNA sequences of these legs were calculated by carrying out a Monte-Carlo optimization using the Java version of DNA Design Toolbox.
Our first approach to design these three legs was based on the previous work done by Lund et al. (Nature, 2010), where they used three legs which consisted of the same sequence. For our purpose we used that sequence as the leg A (figure 3). For leg B and C, unknown sequences have been considered.
Figure 2. The three different DNA-based legs of the molecular robot
The program needs the input of each strand and define which helices exist, from here each leg and its respective field and substrate were defined as a helix. These constrains help us to describe the system in such a way where there is no possibility for the legs to base pair with the robot body.
As legs should not form base pairs with the robot body we include the whole robot body sequence.
Here, we decided to include a second sequence (sticky motif robot), in addition to the prism robot sequence, as a guarantee in case that the prism robot were not successful. Thus, in that case making only the sticky motif structure with the calculated legs.
Unknown values for the nucleotides are represented as N (any base: A, C, G or T) according to the IUPAC notation.
For carrying out the Monte-Carlo optimization, it is needed to set a random number (seed). Unfortunately, there is no relation between how good the optimized sequence is and the given random number that generate it . Therefore, we selected five seeds and found between them the corresponded sequence which generated the lowest best score.
Figure 4. Linkers for the two different molecular robots: a) molecular spider. b) present molecular robot
We included a 'temporary' sequence defined as the sticky part and to be complementary with the base pairing common area between the substrate and leg. This sequence was used for the sake of maintaining a unique melting temperature in all the leg sequence and, consequently, being thermostable sequences. Without this 'trick' the sequence tends to form a low GC content. Then, leading to a low melting temperature (Tm). For calculating the melting temperature of those preliminary single strand DNA sequences we used the online program DINAMelt. You can download those results by clicking here.
Furthermore, in order to connect the DNA body with the DNA legs we decided to use a T-linker instead of the BioTEG//iSp18//iSp18 linker  that binds to the streptavidin protein body. This linker is about 6nm. Therefore, the T-linker is made of 20 thymine nucleotides (figure 4). We believe that our linker has as well flexible properties.
Figure 3. Robot body with the legs