dorsal/arxiv
View SchemaA study of accuracy and precision in oligonucleotide arrays: extracting more signal at large concentrations
| Authors | Felix Naef, Nicholas D. Socci, Marcelo Magnasco |
|---|---|
| Categories | |
| ArXiv ID | physics/0205031 |
| URL | https://arxiv.org/abs/physics/0205031 |
Abstract
Despite the success and popularity of oligonucleotide arrays as a high-throughput technique for measuring mRNA expression levels, quantitative calibration studies have until now been limited. The main reason is that suitable data was not available. However, calibration data recently produced by Affymetrix now permits detailed studies of the intensity dependent sensitivity. Given a certain transcript concentration, it is of particular interest to know whether current analysis methods are capable of detecting differential expression ratios of 2 or higher. Using the calibration data, we demonstrate that while current techniques are capable of detecting changes in the low to mid concentration range, the situation is noticeably worse for high concentrations. In this regime, expression changes as large as 4 fold are severely biased, and changes of 2 are often undetectable. Such effects are mainly the consequence of the sequence specific binding properties of probes, and not the result of optical saturation in the fluorescence measurements. GeneChips are manufactured such that each transcript is probed by a set of sequences with a wide affinity range. We show that this property can be used to design a method capable of reducing the high intensity bias. The idea behind our methods is to transfers the weight of a measurement to a subset of probes with optimal linear response at a given concentration, which can be achieved using local embedding techniques.
{
"annotation_id": "211127bf-2702-44fd-9f2e-85b73c207038",
"date_created": "2026-03-02T18:00:39.718000Z",
"date_modified": "2026-03-02T18:00:39.718000Z",
"file_hash": "c0e4d742abe2dc38ff1a0cce124f5fc24ccdd524d9ef323d23a07d5c633cba4f",
"private": false,
"record": {
"abstract": "Despite the success and popularity of oligonucleotide arrays as a\nhigh-throughput technique for measuring mRNA expression levels, quantitative\ncalibration studies have until now been limited. The main reason is that\nsuitable data was not available. However, calibration data recently produced by\nAffymetrix now permits detailed studies of the intensity dependent sensitivity.\nGiven a certain transcript concentration, it is of particular interest to know\nwhether current analysis methods are capable of detecting differential\nexpression ratios of 2 or higher. Using the calibration data, we demonstrate\nthat while current techniques are capable of detecting changes in the low to\nmid concentration range, the situation is noticeably worse for high\nconcentrations. In this regime, expression changes as large as 4 fold are\nseverely biased, and changes of 2 are often undetectable. Such effects are\nmainly the consequence of the sequence specific binding properties of probes,\nand not the result of optical saturation in the fluorescence measurements.\nGeneChips are manufactured such that each transcript is probed by a set of\nsequences with a wide affinity range. We show that this property can be used to\ndesign a method capable of reducing the high intensity bias. The idea behind\nour methods is to transfers the weight of a measurement to a subset of probes\nwith optimal linear response at a given concentration, which can be achieved\nusing local embedding techniques.",
"arxiv_id": "physics/0205031",
"authors": [
"Felix Naef",
"Nicholas D. Socci",
"Marcelo Magnasco"
],
"categories": [
"physics.bio-ph",
"q-bio"
],
"title": "A study of accuracy and precision in oligonucleotide arrays: extracting more signal at large concentrations",
"url": "https://arxiv.org/abs/physics/0205031"
},
"schema_id": "dorsal/arxiv",
"source": {
"execution_id": "87bc9f84-61a8-426f-8fe4-676643c0f98e",
"id": "arXiv Dataset IDs",
"type": "Model",
"variant": "snapshot-2026-03-01",
"version": "0.1.0"
},
"user_id": 1000002
}