SBIR-STTR Award

This proposal aims to increase the capability and decrease the cost of decoding the Illumina bead array platform by adding decode states. DNA probes attached to microbeads are randomly loaded onto fiber optic bundles. A decoding process of sequential hybridization stages is necessary to determine the locus correspondence of each bead. Decoders (sequences complementary to the DNA probes on the beads) that are either unlabeled, or labeled with a dye are hybridized to the array. Images are taken after each hybridization, and the experiment is designed so that the hybridization signature of each bead through the decode process, uniquely determines the identity of the bead. The cost and time of decoding is proportional to the number of decode stages. The number of stages is determined by the number of loci represented on the array and the number of distinguishable labels, or decode states, used in the decode process (e.g. ON in dye 1). The current availability of 3 states allows the decoding of 1,500 probes in 8 stages. The successful execution of this project would extend the number of states to at least 8. With 8 states, the number of stages for the 1,500-probe product would become 4. The number of probes that could be decoded in 8 stages would increase by 3 orders of magnitude. The main components of the project are wet lab chemistry and algorithm development. Wet lab chemistry will be used to determine the optimal mixture of dye labeled and unlabeled oligonucleotides that will lead to distinguishable intensity states. Beads will have signal levels in FAM and CY3 dye. Variability in the process will need to be sufficiently low to reliably distinguish different concentrations of dye labeled oligonucleo tides. Three levels of FAM and CY3 signal would lead to 9 states. It is likely that 8 of these will be reliably distinguishable. Pattern matching algorithms will be developed to decode the beads. Decision tree methods based on expected signal will be applied. Arrays will be decoded twice -- first with the current 3-state decoding, and then with the multi-state decoding -- to enable training and machine learning algorithms. Achieving 8 state decoding will decrease the cost of the array and dramatically increase the number of loci explored and the number of probes per locus

This proposal aims to increase the information content and decrease the cost of the Illumina random array platform. Random arrays are assembled by attaching DMA probes to silica beads and loading the beads onto an etched substrate. The bead identities are determined through a sequential hybridization based decode process. Once the arrays are decoded, they can be used in applications such as SNP genotyping and gene expression profiling. The cost of decoding is proportional to the number of sequential hybridizations. The number of probes that can be decoded on the array is a function of the number of hybridizations and the number of distinguishable labels (decode states) used in the process. An increase in the number of probes requires either an increase in the number of stages hybridizations or an increase in the number of states. Prior to the completion of Phase I, arrays containing 1,500 probes were decoded with 8 hybridizations and 3 states. The completion of Phase I has demonstrated the feasibility of 7-state decoding. The increase in states dramatically increases decoding capabilities, allowing the decoding of 24,000 probes in 6 stages. The first goal of Phase II is to implement the 7-state decode system in manufacturing. The next goal is to improve and extend the decode system through the addition of intermediate intensity states and the reduction of process variability. The ability to efficiently decode tens of thousand of probes presents an opportunity to increase the density of the random arrays. Arrays with a range of feature sizes will be manufactured and the optimal density will be determined. The efficient decoding combined with the higher density of features will lead to microarrays with high information content, low cost per data point, and low sample consumption