Navigation Links
The New Affymetrix GeneChip Scanner 3000

Seamless Performance Between GeneChip Scanner 3000 and the Affymetrix GeneArray 2500 Scanner


The GeneChip Instrument System is the most advanced microarray analysis platform available and reduces the time and cost of discovery research. Improvements in array manufacturing processes have enabled Affymetrix to produce expression probe arrays with greater consistency and higher overall intensity values. Affymetrix has now taken the next step and developed instrumentation to keep pace with these advancements in probe array technology. As the centerpiece of the newest generation of GeneChip instrumentation, the Affymetrix GeneChip Scanner 3000 is the first step in further revolutionizing microarray analysis platforms, and will seamlessly incorporate future developments as GeneChip technology evolves. Advancements within the Affymetrix GeneChip system will build upon the GeneChip Scanner 3000 and address the requirements of new and emerging applications.

Today, the GeneChip Scanner 3000 incorporates advanced design improvements resulting in:

Space savings
Increased speed
Integrated sample tracking
Robust linearity
Higher resolution
Greater data integrity
Improved assay performance

Many of the technical features and specifications of the GeneChip Scanner 3000 are detailed in the Affymetrix GeneChip Scanner 3000 Specification Sheet.

To ensure the smooth transition of customers to the GeneChip Scanner 3000 scanning system, it is critical that data generated using this scanner are comparable to data generated using the current GeneArray 2500 scanner. In other words, there must be high concordance between experimental data generated from a GeneChip Scanner 3000 and the current production scanne 1.50 pM target spike was similar to that for the 3.00 pM target spike (73-96%), with the GeneArray 2500 at Site 4 generating the lowest average Detection accuracy call. Students t-tests comparing the GeneArray 2500 and GeneChip Scanner 3000 at Site 4 for 1.50 pM indicate that this difference in Detection accuracy is statistically significant (Figure 4, *). However, all values were within specification.

Detection accuracy of 0.75 pM target spikes indicated similar performance for the GeneArray 2500 and GeneChip Scanner 3000 across all sites, with statistically significant performance differences favoring the GeneChip Scanner 3000 at Sites 2 and 4, (Figure 4, *). Percent accuracy of Detection across the sites was 74-87% for the GeneChip Scanner 3000, compared to the range of 54-81% for the GeneArray 2500. Scanner Absent call accuracy for the 0.00 pM target spike was statistically higher for the GeneArray 2500 at Site 4 versus the GeneChip Scanner 3000 (Figure 4, **), with all other scanner comparisons indicating equivalence.

Theoretically, it is expected that without spiked transcripts (0.00 pM), Absent call accuracy should approach 100%. However, a range of 88-95% call accuracy was observed for all scanners (Figure 4, 0.00 GeneChip Scanner 3000 and 0.00 GeneArray 2500). The 0.00 pM target spike represents sample in which the 53 cRNA transcripts were shown to be Absent by previous GeneChip expression assays. An explanation for the apparent discrepancy in the Absent call accuracy is that some of these transcripts may be expressed at a low level and thus, are occasionally detected.

Evaluation of the 53 spiked transcripts showed that a subset of these transcripts was frequently called Present or Marginal in the 0.00 pM target sample across the scanner types and sites. Taqman RT-PCR analysis was performed to examine the possibility of endogenous levels of all 53 transcripts. This analysis showed marginal evidence of expression for essentially the same subset of transcripts (data not shown). This observation may suggest a limitation in the experimental design of the test plan; however, it does not compromise the validity of the conclusion that the scanners perform similarly in most cases.

Accuracy of Target Spike Change Call Comparisons

Change calls reported in the MAS software are indicators of differential gene expression. Obtaining a Change call requires comparison between single probe analyses that may contain different concentrations of certain transcripts. The spiked cRNA target samples permit comparisons that will generate predictable Change calls for these probe sets for both scanner types.

Comparisons across different target spike concentrations were performed on single array analysis (.chp) files generated from scans acquired on either scanner type at each external site: 0.00 pM versus 0.75 pM; 0.00 pM versus 1.50 pM; 0.00 pM versus 3.00 pM; 0.75 pM versus 1.50 pM; 0.75 pM versus 3.00 pM; and 1.50 pM versus 3.0 pM.

Figure 5 illustrates the average percent target spike Increase call accuracy for each comparison at each site using HG-U133A probe arrays. The number of Increase calls was calculated in the MAS software across all 53 spiked transcripts. Average Increase call accuracy was calculated for all possible pair-wise comparisons of the three replicates at two concentrations, resulting in nine comparison analyses.

These results illustrate that, in general, the GeneChip Scanner 3000 and the GeneArray 2500 perform similarly across the range of comparisons, accurately calling the majority of the spiked transcripts as Increases (Figure 5). Statistical analysis showed similar Increase call accuracy results between scanner types, with a potential bias toward the GeneChip Scanner 3000 in some cases (Figure 5, *)

Spiked target concentrations in thi s test plan were deliberately chosen at levels close to the limit of detection. These concentrations were selected to challenge scanner performance with samples that represent low-level expression across a range of concentrations. Thus, certain spiked transcript comparisons appear to result in lower accuracy for both scanner types (Figure 5, 0.75 pM versus 1.50 pM and 0.00 pM versus 0.75 pM). The range of Increase call accuracy for 0.00 pM versus 0.75 pM was 81-91% for the GeneArray 2500, and 85-96% for the GeneChip Scanner 3000. For 0.75 pM versus 1.50 pM, the range was 59-75% for the GeneArray 2500 and 63-77% for the GeneChip Scanner 3000. This observation does not compromise the validity of the conclusion that both scanners perform similarly in most cases.

Accuracy of Target Spike Signal Log Ratio Comparisons

Signal Log Ratio (SLR) is a quantitative measure describing differential expression of a transcript across experimental conditions. It is derived from the one-step Tukeys Biweight method5 for Signal values for a given probe set compared across two experiments. SLR is the log2 of the ratio and is calculated by the MAS analysis software. Therefore, a SLR equal to 1 or -1 is equivalent to a two-fold increase or decrease in expression, respectively.

For comparison analyses between different target spike concentrations, there is an expected SLR for each of the 53 transcripts. For each scanner, the SLR analysis was performed for all concentration comparisons for which a SLR could be predicted.

Figure 6 illustrates the average SLR generated on HG-U133A probe arrays for target spike concentrations where pair-wise comparisons result in a SLR equal to 1 (0.75 pM versus 1.50 pM and 1.50 pM versus 3.00 pM). The SLR shown was calculated for all possible pair-wise comparisons of three replicates at two concentrations resulting in nine comparison analyses. Error bars represent one standard deviati on. The data for all SLR comparisons showed equivalence in the GeneChip Scanner 3000 and GeneArray 2500 performance.

The comparisons in Figure 6 are the most challenging because the target spike concentrations of 0.75 pM and 1.50 pM are close to the limit of detection of the assay. The average SLR ranges for 0.75 pM versus 1.50 pM were 0.61-0.70 and 0.57-0.69 for the GeneChip Scanner 3000 and GeneArray 2500, respectively. The average SLR ranges for 1.50 pM versus 3.00 pM were 0.79-0.85 and 0.81-0.84 for the GeneChip Scanner 3000 and GeneArray 2500, respectively. These values of


The results for human genome probe arrays described in this Technical Note clearly indicate that the performance of the GeneChip Scanner 3000 is equivalent to or better than the performance of the GeneArray 2500. Where statistical analysis was applied, Students t-tests provided substantial evidence of performance equivalence of the two scanner types. In the subset of cases where a bias may have been indicated, most favored the performance of the GeneChip Scanner 3000.

At Affymetrix, we take pride in the high quality of products we develop and support. We are pleased to Conclusion AFFYMETRIX UK Ltd Voyager, Mercury Park, Wycombe Lane, Wooburn Green, High Wycombe HP10 0HH United Kingdom Tel: +44 (0)1628 552550 Fax: +44 (0)1628 552585 AFFYMETRIX, INC. 3380 Central Expressway Santa Clara, CA 95051 USA Tel: 1-888-362-2447 (1-888-DNA-CHIP) Fax: 1-408-731-5441 For research use only. Not for use in diagnostic procedures. Part No. 701218 Rev. 1 2003 Affymetrix, Inc. All rights reserved. Affymetrix,GeneChip, , , , HuSNP, Jaguar, EASI, MicroDB, GenFlex , CustomExpress, NetAffx, CustomSeq, Tools to take you as far as your vision, and The Way Ahead are trademarks owned or used by Affymetrix, Inc. Products may be covered by one or more of the following patents: U.S. Patent Nos. 5,834,758; 5,936,324; 5,981,956; 6,025,601; 6,141,096; 6,185,030; 6,201,639; 6,207,960; 6,218,803; 6,225,625; 6,252,236; 6,335,824; 6,403,320; 6,490,533; and other U.S. or foreign patents. introduce an advanced scanning system with the Affymetrix GeneChip Scanner 3000. As this Technical Note supports, current GeneChip users may easily transition to the GeneChip Scanner 3000 with no sacrifice to their current experimental designs and projects. In addition, the use of the GeneChip Scanner 3000 will position Affymetrix customers to take full advantage of upcoming technological advancements within the Affymetrix GeneChip platform and address the needs and requirements of new, emerging applications.

For further information, please refer to the Affymetrix GeneChip Scanner 3000 Specification Sheet.

For additional questions and purchase information, in North America, please call 1-888-DNA-CHIP or contact your local Account Manager. In other regions, please contact your local Affymetrix representative.

1 New Statistical Algorithms for Monitoring Gene Expression on GeneChip Probe Arrays (Affymetrix Technical Note), Part No. 701097 (2001).

2 Statistical Algorithms Reference Guide (Affymetrix Technical Note), Part No. 701110 (2001).

3 GeneChip Expression Analysis: Data Analysis Fundamentals (Affymetrix publication), Part No. 701190 (2002).

4 A.R. Feinstein, "Clinical Biostatistics: A Survey of the Statistical Procedures in General Medical Journals," Clinical Pharmacological Therapy , 15: 97-107 (1974).

5 D.C. Hoaglin, F. Mosteller, and J. Tukey, Understanding Robust and Exploratory Data Analysis, New York: John W iley & Sons, (2000).

r, the GeneArray 2500.

This Technical Note details a set of controlled comparison experiments between the new GeneChip Scanner 3000 and the existing GeneArray 2500. The results demonstrate conclusively that the GeneChip Scanner 3000 meets or exceeds performance standards for the GeneArray 2500 in multiple independent comparisons.

Experimental Design


The newly designed GeneChip Scanner 3000 was developed to perform equivalently to or better than the existing GeneArray 2500. To evaluate the GeneChip Scanner 3000, two comprehensive test plans were executed to compare the GeneChip Scanner 3000 to the GeneArray 2500. These test plans were similar in that they shared common objectives regarding instrument performance:

Identify differences in global GeneChip probe array metrics, such as percent Present Calls, False Change between replicates, and correlation coefficients from Signal values. (Detailed in this Technical Note.)
Identify differences in the ability of the scanners to detect and quantify exogenous spiked transcripts in a complex cRNA target background, considering metrics such as Detection call accuracy, Change call accuracy, and Signal Log Ratio accuracy. (Detailed in this Technical Note.)
Survey customer satisfaction for the GeneChip Scanner 3000. (Detailed in the Affymetrix GeneChip Scanner 3000 Specification Sheet and other marketing materials.)

Both test plans involved acquiring data with both the GeneChip Scanner 3000 and GeneArray 2500 using the same GeneChip probe array scanned in both instruments. Any differences observed were due exclusively to the scanners. All analyses were conducted with the Affymetrix Microarray Suite 5.0 (MAS) software1, 2, 3.

Beta Test Plan

test plan compared the performance of six pairs of scann ers (one GeneChip Scanner 3000 and one GeneArray 2500) using Affymetrix GeneChip human genome probe arrays (HG-U133A and HG-U95Av2). Six customer sites were chosen to represent academia, biotech, and pharmaceutical companies of varying GeneChip probe array use and experience. Standard preventive maintenance was performed on all GeneArray 2500 scanners prior to the test, ensuring the scanners were within specifications.

The beta test sites were provided probe arrays, four hybridization targets containing spiked transcripts (at four concentrations, discussed later in the document), staining reagents, a GeneChip Scanner 3000 with experimental instructions, and support from Affymetrix. Beta test participants were responsible for acquiring images on both the GeneChip Scanner 3000 and GeneArray 2500, returning the raw data to Affymetrix, and providing detailed feedback on the GeneChip Scanner 3000.

Hybridization targets (labeled and fragmented cRNA ready for hybridization) were prepared from human cell line total RNA following procedures outlined in the GeneChip Expression Analysis Technical Manual.

Assay performance was evaluated, in part, on the ability of each scanner type to detect and quantify 53 labeled cRNA transcripts spiked into the target sample. These 53 transcripts were selected on the basis of their apparent lack of expression in the source RNA, as determined by their Absent calls on GeneChip human genome probe arrays.

The spiked transcripts were generated by in vitro transcription from vectors containing the cloned DNA sequence. Labeled cRNA transcripts were then quantified and fragmented and spiked back into labeled, fragmented, complex background cRNA, to reconstitute expression of all 53 transcripts at one of three concentrations: 0.75 pM (picomolar), 1.50 pM, or 3.00 pM. A target sample created from background cRNA without spiked tra nscripts added was considered a fourth concentration of 0.00 pM. Large pools of complex sample containing the 53 transcripts at each of the four concentrations were created and split among the probe arrays; thus, identical samples were run on all scanners and probe array designs.

Two human genome probe array designs (HG-U95Av2 and HG-U133A) were used with three lots of probe arrays per design. Samples were run once on each array. Therefore, each external site generated 48 scans: 2 probe array designs x 3 lots of each probe array design x 1 sample for each lot x 4 spiked target samples x 2 scanners (Table 1).

Scanning Strategy

Each probe array was scanned first on the GeneChip Scanner 3000, then on the GeneArray 2500. This scanning strategy permitted scanning a single probe array per pair of instruments, eliminating variation introduced by different probe arrays, and ensuring that any differences in performance were attributable to the scanner that generated the image.

Photobleaching that may occur due to scanning in the GeneChip Scanner 3000 prior to scanning in the GeneArray 2500 is negligible, and has been shown to have no impact on image quality or the resultant data (data not shown). Furthermore, the scanning order between the GeneChip Scanner 3000 and GeneArray 2500 was reversed in limited tests and showed no impact on results (data not shown).

Alpha Test Plan

An initial alpha test plan was conducted at Affymetrix and compared the performance of three pairs of scanners (one GeneChip Scanner 3000 and one GeneArray 2500). The alpha test plan had the same experimental design, probe arrays, supplies, and reagents used for the beta test plan. However, samples were run in triplicate on each array lot, therefore generating 144 internal site scans: 2 probe array designs x 3 lots of each probe array design x 3 replicate samples for each lot x 4 spi ked target samples x 2 scanners.

Performance Data & Analysis

In both the internal and external test plans, the performance of each scanner was assessed on the basis of both single array and comparison analyses generated within the MAS software. Analysis output files (.chp files) were created at Affymetrix by globally scaling all probe sets on image data files (.dat files) to a target intensity of 250. Data sets were generated to assess overall Signal correlation, global percent Present calls, and False Change between replicates. In addition, spiked transcript Detection call accuracy, Change call accuracy, and Signal Log Ratio accuracy between scanner types were evaluated.

Data from each set of scanners were tested for statistical significance, where appropriate, by applying paired Students t-tests4 for both the HG-U95Av2 and HG-U133A probe arrays. The null hypothesis was defined as the GeneArray 2500 performance significantly exceeded the GeneChip Scanner 3000 performance at a p>0.95 confidence level. Therefore, p-values less than or equal to (≤) 0.95 were interpreted as contradicting the null hypothesis; p-values approaching 0.5 were interpreted as equivalent scanner performance; and p-values <0.05 were interpreted as significantly better performance of the GeneChip Scanner 3000.

Results & Discussion

Results Summary

Taken together, all array data comparing the GeneChip Scanner 3000 to the GeneArray 2500, from both internal alpha testing and external beta testing, indicate that there is no significant performance difference between the two scanner types. Both scanner types produce high-quality data for Affymetrix GeneChip expression analysis.

Results from both HG-U95Av2 and HG-U133A probe arrays were independent of the scann er used to generate the data or the site at which the comparison was made. Array data produced at each site yielded statistical equivalence in most cases.

Students t-test4 p-values obtained for all analyses (for example, percent Present, Detection call accuracy, etc.) tended to distribute about 0.5, with perhaps a slight bias in favor of superior performance by the GeneChip Scanner 3000. Instances where scanner performance was different are noted in the following results.

Due to the equivalence of results between alpha and beta tests, the results described in this Technical Note are taken from the beta external tests. In addition, because conclusions reached were the same for both human probe array designs, only the HG-U133A probe array data is discussed in detail. Beta sites that deviated from the established test plan were not included in the analyses. Therefore, results from four of the six beta test sites are described.

Signal Correlation Between Scanner Types

Scatter plots of scaled Signal values for all transcripts show a high degree of similarity between scanner types and across external sites.

For this analysis, a data set from a hybridized target sample was selected as representative of all HG-U133A probe arrays. Signal values for probe sets from this data set were directly compared to the Signal values from the same probe sets at all other sites and on both scanner types. Thus, all 22,283 probe sets from the HG-U133A probe array and their Signal values were used for each correlation coefficient determination.

Correlation coefficients calculated for each scanner comparison, chip design, and site are summarized in Table 2. The high degree of similarity of Signal values between the GeneChip Scanner 3000 and GeneArray 2500 within each site is indicated by correlation coefficients approaching 1.0. The corr elation coefficients for other Signal comparisons (GeneChip Scanner 3000 versus GeneChip Scanner3000 and GeneArray 2500 versus GeneArray 2500 at different sites) were also as high as those listed in Table 2 (data not shown).

Figure 1 shows scatter plots generated for these analyses. The high correlation of Signal values generated by the GeneChip Scanner 3000 and the GeneArray 2500 at the same and different external sites is illustrated by plots A and B, respectively.

Percent Present Comparisons

The percent Present metric can be a useful measure of assay and instrument performance. It describes the total number of transcripts detected (called Present) out of the total number of probe sets on a probe array.

Figure 2 illustrates the average percent Present for 12 HG-U133A arrays run on each scanner type at each external site. Error bars represent one standard deviation.

The GeneChip Scanner 3000 average percent Present is statistically different from the GeneArray 2500, with the GeneChip Scanner 3000 yielding a higher average percent Present call for probe arrays run at all sites (Figure 2, *). For probe arrays run at Site 4, the GeneArray 2500 yielded a noticeably lower percent Present when compared to the GeneChip Scanner 3000 at the same site and GeneArray 2500 scanners at other sites (Figure 2, *). While the percent Present metric of the GeneArray 2500 at Site 4 appears to be different than other GeneArray 2500 scanners, it is within Affymetrix specifications and is representative of the range of GeneArray 2500 scanner performance across Affymetrix customer sites.

False Change Comparisons

False Change is a metric used to characterize reproducibility of results when scanning the same target sample hybridized to different probe arrays of the same design. False Change is defined as the frequency of the sum of Increa se calls greater than or equal to (≥) a Signal Log Ratio of 1 and Decrease calls less than or equal to (≤) a Signal Log Ratio of 1 for a comparison analysis between identical replicate samples. When the same probe array is scanned on two different scanner types, this metric is also a useful measure of scanner consistency.

Figure 3 illustrates the average False Change for sample targets hybridized to three replicate HG-U133A arrays across the different spiked target concentrations at each site. Therefore, for each scanner type at each site, the False Change listed is an average from 12 samples. Error bars represent one standard deviation.

The specification for HG-U133 probe arrays is a False Change of less than 1%. The average False Change across all sites is well below 1% and is statistically indistinguishable between the GeneChip Scanner 3000 and GeneArray 2500. A possible explanation for the slight variation in False Change across the sites (for example, Figure 3, Site 1 versus Site 2) is due to differences in probe array processing, including hybridizing probe arrays on different days, using different operators, using different fluidics stations and wash buffers, etc.

Accuracy of Target Spike Detection Comparisons

The results discussed previously utilized the entire probe set population for analysis and thus, were considered global metrics measuring instrument performance. To further compare instrument performance, additional analyses (discussed next) focus on the cRNA spikes described for the alpha and beta test plans.

The use of target spikes at concentrations of 0.00 pM, 0.75 pM, 1.50 pM, and 3.00 pM in a complex biological material provided a calibrated measure of instrument performance at the limits of detection of the system. The transcripts at 1.50 pM correlate to a frequency of 1 transcript in 100,000, and are considered a reliable measu re of GeneChip instrument system sensitivity (greater than or equal to 70% of the time these transcripts are detected).

Figure 4 illustrates average Detection call accuracy for the 53 transcripts at each spiked concentration. Target spike concentrations correspond to sample targets containing the 53 transcripts presumed to be Absent (0.00 pM) based on historical results from GeneChip HG-U133 expression assay analysis of this HeLa cell line, and then reconstituted to 0.75 pM, 1.50 pM, or 3.00 pM in the final Figure 3. Comparison of Average Percent False Change Between GeneChip Scanner 3000 and GeneArray 2500 at External Sites. False Change is described by the frequency of Increase calls greater than or equal to a Signal Log Ratio of 1 and Decrease calls less than or equal to a Signal Log Ratio of -1 for a comparison analysis between identical replicate samples hybridized to two probe arrays of the same design. Each histogram shown reflects the average percent False Change for 12 target samples hybridized to three lots of HG-U133A probe arrays scanned on a particular scanner. Error bars represent one standard deviation. See text and Figure 2 for sample description. 1.8% 1.6% 1.4% 1.2% 1% 0.8% 0.6% 0.4% 0.2% 0% Site 1 Site 2 GeneArray GeneChip Scanner 3000 y 2500 Site 3 Site 4 Site % False Change (Average) False Change Comparisons Between Scanner Types complex cRNA target. The histogram value is the average frequency for which a Present Call is made for the 53 transcripts at 0.75 pM, 1.50 pM, or 3.00 pM; or the average frequency for which an Absent call is made for the 0.00 pM target. For each scanner type at each site, the percent accuracy of Detection is an average of three hybridized target samples. In general, the analysis shows that accurate transcript detection is similar for both scanner types at most sites.

The 3.00 pM target spike was called Present with nearly 100% fidelity by all scanners at all sites. Detection accuracy for the


Page: All 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15

Related biology technology :

1. Fine Tuning Your Data Analysis: Tunable Parameters of the Affymetrix Expression Analysis Statistical Algorithms
2. The Affymetrix GeneChip Scanner 3000 New Advances in Scanner Design for Superior Performance, Reliability, and Dynamic Range
3. GeneChip Sample Cleanup Module for cleanup of cDNA and in vitro transcribed cRNA
4. New statistical algorithms for Monitoring Gene Expression on GeneChip® Probe Arrays
Post Your Comments:

(Date:11/25/2015)... Nov. 25, 2015  Neurocrine Biosciences, Inc. (Nasdaq: ... President and CEO of Neurocrine Biosciences, will be presenting ... New York . ... website approximately 5 minutes prior to the presentation to ... the presentation will be available on the website approximately ...
(Date:11/25/2015)... ... November 25, 2015 , ... Jessica ... on AngelList early in their initial angel funding process. Now, they are paying ... looking to make early stage investments in the microbiome space. In this, ...
(Date:11/24/2015)... , Nov. 24, 2015 /CNW/ - iCo Therapeutics ... today reported financial results for the quarter ended ... expressed in Canadian dollars and presented under International ... United States ," said Andrew Rae ... advancements regarding iCo-008 are not only value enriching ...
(Date:11/24/2015)... India , November 24, 2015 ... a new market research report "Oligonucleotide Synthesis Market by ... Application (PCR, Gene Synthesis, Diagnostic, DNA, RNAi), End-User (Research, ... 2020", published by MarketsandMarkets, the market is expected to ... Million in 2015, at a CAGR of 10.1% during ...
Breaking Biology Technology:
(Date:11/2/2015)... PARK, Calif. , Nov. 2, 2015  SRI ... $9 million to provide preclinical development services to the ... the contract, SRI will provide scientific expertise, modern testing ... wide variety of preclinical pharmacology and toxicology studies to ... --> The PREVENT Cancer Drug Development ...
(Date:10/29/2015)... 2015  The J. Craig Venter Institute (JCVI) policy ... and Biosecurity: Lessons Learned and Options for the Future," ... Human Services guidance for synthetic biology providers has worked ... --> --> Synthetic biology promises ... to pose unique biosecurity threats. It now is easier ...
(Date:10/29/2015)... Mich. , Oct. 29, 2015  Rubicon ... Genomics for U.S. distribution of its DNA library ... kit and Rubicon,s new ThruPLEX Plasma-seq kit. ThruPLEX ... enable the preparation of NGS libraries for liquid ... for diagnostic and prognostic applications in cancer and ...
Breaking Biology News(10 mins):