Tabu search-based synthesis of dynamically reconfigurable digital microfluidic biochips

Abstract

Microfluidic biochips are replacing the conventional biochemical analyzers, and are able to integrate on-chip all the necessary functions for biochemical analysis. The "digital" microfluidic biochips are manipulating liquids not as a continuous flow, but as discrete droplets, and hence they are highly reconfigurable and scalable. A digital biochip is composed of a two-dimensional array of cells, together with reservoirs for storing the samples and reagents. Several adjacent cells are dynamically grouped to form a virtual device, on which operations are executed. During the execution of an operation, the virtual device can be reconfigured to occupy a different group of cells on the array. In this paper, we present a Tabu Search metaheuristic for the synthesis of digital microfluidic biochips, which, starting from a biochemical application and a given biochip architecture, determines the allocation, resource binding, scheduling and placement of the operations in the application. In our approach, we consider moving the modules during their operation, in order to improve the completion time of the biochemical application. The proposed heuristic has been evaluated using three real-life case studies and ten synthetic benchmarks.

Fast Online Synthesis of Digital Microfluidic Biochips

“…MLS produces schedules comparable to GA in much less time. Other scheduling algorithms such as Ricketts' hybrid genetic algorithm [21] and Maftei's tabu search scheduler [16] are iterative improvement algorithms which spend anywhere from four seconds to one hour computing schedules. We chose list scheduling as the base scheduler for our framework, but other fast schedulers being developed now [11], [18], or in the future could be used as well.…”

“…LS is a greedy, constructive algorithm in which each operation (node) in an assay (DAG) is scheduled exactly once. LS is much faster than iterative improvement algorithms, which randomly compute numerous schedules [16], [21], [26] or optimal algorithms based on integer linear programming [26]; however, these approaches generally do produce higher quality schedules than LS. PS is another scheduler that attempts to schedule DAGs one path at a time (as opposed to a single node at a time).…”

“…If n has multiple children (e.g., split operation), then only the first eligible child (randomly selected) is added to the current path node, p; a new path node is created for each remaining eligible and ineligible child and inserted into G p and the path leaders list (lines [14][15][16][17][18][19]. Similarly, if n only has one unvisited child, the child is either added to the current path node, p loop through and choose a module location for each storage operation, sn.…”

“…In this section, we highlight the key benefits of the virtual topology and binder by comparing Path Binding to a fast free placement algorithm known as Keep All Maximal Empty Rectangles (KAMER) placement [4], [16]. The KAMER placer works by quickly computing all the maximal empty rectangles (MERs) (i.e., the empty rectangles that cannot be contained within another empty rectangle) and then placing a module within one of the MERs.…”

Force-Directed List Scheduling for Digital Microfluidic Biochips

“…Yu et al [14] combine scheduling and placement into a 3D placement problem in space and time; they solve the problem using simulated annealing and a data structure (the T-tree) to represent the placement. Maftei et al [6] solve scheduling, module selection, placement, and routing using Tabu Search. These iterative improvement algorithms are much like GA-1 and GA-2 in terms of quality and runtime.…”