Biomedical Engineering Reference
In-Depth Information
noise from magnetic fields from nearby wiring and devices. There are also short-term errors due to
equipment warm-up, and long-term drift with time because of aging of the electronic components.
Variation can be introduced by instrument loading and perturbation of the system under study, or by
crosstalk. Crosstalk occurs when, for example, light from one channel or florescent color bleeds over
to a detector intended for another signal. Using emission filters that block potential interfering light
signals can minimize it. Variability can be introduced by noise in the power supply, by memory
effects (image persistence) on the image sensors due to previous exposures, and by drift of sensor
sensitivity and amplifier gain with time. Simply using the measuring equipment as it's intended to be
used can demonstrate susceptibility to errors. For example, there is the issue of instrument loading,
in that following Heisenberg's Uncertainty Principle, it's generally impossible to measure something
with absolute accuracy without changing its value to some degree. For example, because of
photobleaching by the laser, the process of reading a microarray also erases data from the
microarray, making subsequent readings less accurate.
Imperfections in measuring equipment can introduce variation in the data. For example, because
insulation used on wires is imperfect, there is current leakage and equipment noise. In addition,
spurious signals can be induced by mechanical stress on wires and electronic components (the
Piezoelectric effect), by friction, such as when materials rub together (the Triboelectric effect), or
when insulation quality changes due to high humidity or because of surface contamination. Noise can
also be induced by current-carrying cables and wires located near the measuring equipment.
Deciding on components to use in constructing a microarray system or other complex measuring
system is a compromise between price, performance, and the intended use. For example, ceramic
insulators have a high volume resistance but compared to cheaper polyvinyl chloride (PVC)
insulators, they are a source of noise at high humidity and when subject to physical stress. Noise can
also be produced by electrochemical effects on the circuit board and through thermoelectric
potentials induced by conductors of different composition touching each other. Although these and
other sources of low-level noise may not be relevant in typical bioinformatics work, it's important to
realize the spectrum of possible sources of variation that, taken together, can affect the data
produced by microarray experiments.
