Lab-on-a-chip (LoC), a kind of biochip, is one of the most popular research topics in recent years [1]. An LoC is an analysis system that integrates many biochemical functions in a small chip, such as injection, mixing, separation, and detection [2]. Compared with conventional biochemical systems in labs, hospitals, or research centers, which are always bulky and expensive, LoCs offer many advantages like portability, reagent volume reduction, automation, mass production, fast analysis, high throughput, and low power consumption [3].
As the technology has been advancing in recent years, a new type of LoC, Digital Microfluidic Biochip (DMFB), has been developed. Figure 1(a) illustrates a real DMFB design, and Figure 1(b) shows a typical structure based on 2D array. A DMFB can be used to carry out various bioassays by precisely controlling some discrete and small volume fluidic droplets containing biochemical samples or reagents. Since all the reactants are dispensed as discrete droplet, we called this kind of biochip “digital” one.
(a) A real case of DMFB [6].
Electrodes (2D array)
Droplet Detector
Dispensing port (b) A typical 2D array-based structure of DMFBs.
Figure 1. Digital Microfluidic Biochip.
2
The electrowetting-on-dielectrics (EWOD) effect is utilized as an electrostatic actuation method to dispense, transport, split, merge, and mix droplets on DMFBs. Through applying control voltage on the electrodes below the chip, the surface tension of droplets can be changed. The induced force is used to move these droplets, as shown in Figure 2 [4]–[7].
Therefore, a biochemical assay can be conducted via a series of basic droplet operations.
High voltage dropelt
(a) Top view.
Glass
Glass
Droplet
High voltage
(b) Side view.
Figure 2. Electrowetting-on-dielectrics (EWOD) effect.
Recently, lots of on-chip laboratory procedures, such as immunoassay, protein crystallization, and DNA sequencing, have been successfully demonstrated on DMFBs [8].
Because the demand for various applications continuously grows, the design complexity of DMFBs is certainly increased. Therefore, a series of automation algorithms are necessary for speeding up the process, reducing the manual effort, and improving the design quality. In the past few years, lots of design automation algorithms are proposed to tackle problems in DMFB design flow, such as synthesis, placement, routing, control pin assignment, and chip testing [9]–[20]. Undoubtedly, the design automation for DMFB is one of the most emerging topics nowadays, and the related research works are proliferating.
Sample preparation problem is one of the critical issues in DMFB design automation.
Reactants (sample or reagent) must be diluted to the specific concentration, which is called target concentration, in the process. There are some factors may affect the quality of whole dilution process, they are:
(1) Usage of valuable reactants: it is the major cost for a biochemical reaction. The usage of valuable reactant which is very expensive or is limited in amount, like costly reagents or infant’s blood, should be minimized. An analysis would fail if the preparation process does not consider reactant minimization.
(2) Number of waste droplets: since the number of waste reservoirs on a given DMFB is fixed, the capacity for waste handling is accordingly limited. The excessive waste count may lead to long preparation time. Furthermore, keeping too much waste droplets on DMFB is not propitious for droplet routing. That is, it makes the routing more complicated and may require much extra transportation time to drive waste droplets into reservoirs.
(3) Number of dilution operations: it basically represents the preparation time, and thus should be minimized. A long preparation process may be a fatal in some urgent clinical incidents.
Several algorithms which tackle the sample preparation problem on DMFBs have been proposed [24]–[31]. Most of them address on single-target sample preparation. That is, they generate each target CV through an individual dilution process. If multiple target CVs are required, different target concentrations must be produced one-by-one and is thus time-consuming. Moreover, since those methods do not consider reactant sharing among different dilution process, they may lead to higher reactant usage and waste count.
Intermediate droplet sharing algorithm (IDSA) [29] is the first work that deals with the concurrent preparation for multiple target concentrations. It reduces waste amount by means of minimizing the number of intermediate concentrations in dilution process. However, this strategy does not always lead to a better result. Furthermore, in our opinion, minimizing the usage of valuable reactant is as important as waste reduction. Thus, an approach which can achieve both reactant and waste minimization in multi-target sample preparation is necessary.
In this thesis, we propose a waste recycling algorithm (WARA) for multi-target sample preparation on DMFBs. WARA adopts a tree-based dilution strategy. It generates a mixing tree for each target concentration at beginning, and recycles the waste droplets among those trees by droplet sharing and droplet replacement. Experimental results show that WARA reduces the amount of (waste, operations) by (48%, 37%) on average as compared to IDSA if the number of targets is ten1. The reduction can be up to (97%, 73%) as the number of target concentrations grows to 100. The results suggest that WARA should be a better solution for multi-target sample preparation on DMFBs.
The rest of this thesis is organized as follows. Chapter 2 describes the sample preparation process. Chapter 3 briefly introduces previous works. In Chapter 4, we identify our motivation and the problem formulation. Our multi-target sample preparation algorithm,
1 Since IDSA reported its results by means of bar charts, we have done our best to compare our work with IDSA as accurate as possible.
4
WARA, is elaborated in Chapter 5. The experimental results are reported and discussed in Chapter 6. Finally, Chapter 7 concludes this paper.