Home    ||    About DOC    ||    Hardware    ||    Software    ||    Applications & Demonstrations    ||    Data Access Portal   ||  Methods    ||    References & Papers

Methods

Microarray Expression Standards, necessary to reliably compare experimental data

To integrate microarray data, two technologies are required, one being software that can merge data from different array designs (different spotted arrays and Affymetric GeneChips) and that solution is ARROGANT described above.  In addition, since spotted microarrays and certain DOC experiments (alternative splicing, for example) require a reference which was constructed and is in use.  We worked with Stratagene to construct such a reference and evaluate its efficacy.  One reference method uses genomic DNA as a standard, since it is by definition complete and universally available. The most widely used standard employs a highly representative total RNA pool (Stratagene, La Jolla, CA). To determine the advantages and disadvantages of both, these two standards were directly compared by hybridization to UTSW’s 4,000 or 10,800-member spotted human cDNA array. The labeled analytes were 2.0 micrograms of normal human genomic DNA labeled by nick translation, or 20 micrograms of the Stratagene total RNA pool labeled by reverse transcription

(A) The full curve of the log base 10 of the intensity of the spots versus the percentage of spots shows the minimal difference between the two reference standards. The gray line represents the genomic DNA labeled in Cy 5 and the black line represents the Stratagene RNA Pool labeled in Cy 3. (B inset) A blowup of the low intensity threshold region of the curve to better illustrate the differences between the two standards. The flatting of the curve and the length of the tails are measures of the number of spots above threshold and the reproducibility of the standard respectively. The line at the bottom at 1.5 log 10 intensity is where neither standard is above threshold in either channel, and therefore is a measure of coverage failure.

In another published project we investigated the source of array member cross-hybridization.  This research indicated that not only family members, but also many genes that share significant homology (~65 hydrogen bond pairs in length) via certain motifs, especially repeat expansions (that were identified using POMPOUS/Rep-X) also were susceptible to cross-hybridization leading to false results.  These results provided critical input into our DOC chip code to mask off regions potentially probematic due to cross hybridization potential.

Methods and chemistry developments and their implication to the project and other array applications

Probe Design:  Only a small fraction (usually <5%) of all possible short oligonucleotide probes bind efficiently to complementary segments in long RNA transcripts. As a result, judicious selection of oligonucleotide probes is crucial to the success of transcript profiling experiments. To take advantage of custom arrays, rapid computer aided probe selection is required. To develop an algorithm for prioritizing selection of probes, we have analyzed predicted thermodynamic parameters for the binding of several large sets of probes to complementary RNA transcripts and experimentally confirmed their validity. In addition, we have used DOC to generate two new arrays of surface-bound probes and measured the binding of these probes to their RNA targets, transcripts comprising the messages for the tumor suppressor p53 (1162 unique probes) and green fluorescent protein (755 unique probes). We considered predicted free energies for intramolecular base-pairing of the oligonucleotide and its RNA target as well as the predicted free energy of intermolecular hybridization of probe and target. We have created software tools for automated calculation and ranking of the relevant parameters.

The intensity of a set of 1162 probes tilled in 1 base increments across the p53 transcript illustrates the level of non-uniformity of hybridization, and thus emphasizes the need to properly select probes for expression analysis to obtain the highest sensitivity.  Note, this is not an option for re-sequencing, where probes that tile the region are necessary, but in that case, DNA is usually available in quantity.

            The design (selection) of a set of probes on our DOC chips uses a variety of different algorithmic modules, depending on the function of the array (expression, resequencing, methylation analysis, etc.).   Those modules feed an oligonucleotide design module that has different methods available to optimize the probe design.  As an example, the process for a small genome, Moraxella, was to input a fasta format file that contains the DNA sequence and the annotation file into a Perl program that parses the positions of the ORFs from the sequence.  Note, we make DOC chips for collaborators in areas outside cancer research, at their expense, especially for genomes that are available electronically, and for which there are no available reagents.  In this case we downloaded the Moraxella genome from the patent database.  A tiling path of every possible 21mer is computed for each ORF sequence and the free energy of the oligonucleotide sequence is computed. 15 non-overlapping oligos with minimum free energy are selected as the probes for the microarray.  Finally, all the probes that are designed are output in a text file (up to 196,000) which is input into the DOC machine operations code to convert the sequences into a series of tiff image files that are projected in series and in synchrony with the introduction of the photo-protected phosphoamidites during the synthesis process.  The end result is a chip with 15 probes for each of the 1511 ORFs or 22,665 probes.  After hybridization, the signal from the 15 probes are averaged to establish the expression level for each ORF.  For a large genome, for example human, the selection of which database sequence for a given gene is important (mRNA, genomic, Refseq, probes on an Affymetrix chip, commercial olignucleotide sets, etc.), for the different database have different levels of error (single pass vs. high depth sequencing) and may include only one splice version.  For transcript profiling, the gene coding region sequences are assembled and from there a tiling path of every possible 21mer is generated, with the remaining steps the same as for a small genome.  This technique was used to create the chip used in experiments detailed below.  Very recently we have ported our free energy design code onto our 32 CPU Linux cluster to accelerate the design process.

Array Quality Control tools and techniques:  To maintain a high quality of operation, we have implemented a repertoire of experimental protocols characterize the performance of the instrument. One set of tools relies on reaction of a phosphoramidite derivative of the fluorophore Cy3 to quantitatively probe the density of free hydroxyls on the chip surface. In this manner we are able to measure the critical parameters of phosphoramidite coupling yield, stepwise synthesis yield, density of sites of synthesis initiation, rate of photodeprotection, and yield of photodeprotection. Another set of methods qualifies the arrays after synthesis. A series of control features built into each array and labeled oligonucleotide standards co-hybridized with each sample allows quantification of chip performance for efficiency and specificity of hybridization.

Surface Chemistries and photoprotected phosphoramidites:  DOC system development demanded that we implement reliable chemistry for the preparation chip substrates.  Three different silane coatings were tested for free hydroxyl density, phosphoramidite coupling percentage, and overall stepwise yield of an 8mer oligonucleotide. The three silanes were N-3-(triethoxysilyl)-propyl)-4-hydroxybutyramide, bis(2-hydroxyethyl)aminopropyltriethoxysilane, and 3-glycidoxypropyltrimethoxysilane followed by polyethlyeneglycol (M_a 300).  Glycidoxypropyltrimethoxysilane yielded the highest free hydroxyl density and most consistent synthesis yield.  The coupling percentage with the glycidoxypropyltrimethoxysilane is above 98% for all four bases and the stepwise yield was measured at 96%.  This is equal to or better than that obtained by Affymetrix in their industrial process, thus enabling the manufacture of longer probes at a higher final yield.

The stepwise yield is dependent on two variables: the coupling yield and the deprotection yield.  The coupling yield for phosphoramidite synthesis, either standard or light directed synthesis, is ~99%.  However, the removal of the 5’ protection group on the growing stands has been substantially lower in light directed synthesis when compared to standard chemistry. 5’-(a-methyl-2-nitropiperonyl)oxycarbonyl (MenPOC) is the protective group used in the Affymetrix process and has been reported to have a stepwise yield of 92% dry and 96% in the presence of 1,4, dioxane. The stepwise yield of our instrument has been determined in the presence of 1,4 dioxane to be ~96% using MenPOC as a protecting group.  A report using 2-(2-nitrophenyl)-propoxycarbonyl (NPPOC) as a 5’ protecting group has claimed a higher stepwise yield than obtained with MenPOC. Initial experiments tested the performance and overall effects on microarray quality of these new photolabile groups (available commercially from Proligo Ft. Collins CO) on our system.  Currently, coupling greater than 98% on the DOC system is possible, and new deprotection protocols are being tested.