MS: ✗
BD: ✗

Noise Considerations

Our modeled solution will be affected by random noise, time-invariant noise, the available signal, and its ratio to the random noise. How then can we deal with the uncertainty of our solutions resolution?

First, we can improve the signal to noise ratio. This can be done by:

1. Obtaining lots of high-quality data

2. Exploiting the entire dynamic range of the acquisition system

3. Reducing noise and only using a well-functioning instrument

4. removing systematic noise from experimental data

Additionally, we can:

• Replace fitting parameters with experimentally determined values from separate experiments

• Explore the parameter space with a grid method, then parsimoniously regularize solution with GA, and use Monte Carlo to explore confidence regions

• Perform global fits for multiple experimental conditions to improve signal

There are three ways to reduce or eliminate noise to reduce the number of solutions.

1. Fit the noise.

2. Maintain an exceptionally well-tuned instrument.

3. Design your experiment to optimize the quality of data.

Types of Noise

There are at least three types of noise to consider when performing SV-AUC experiments:¹

1. Time-invariant noise which produces a constant offset at a fixed radial position, and is identically valued at every observation. This type of noise is caused by imperfections in the optical track; for example, a fingerprint or scratch on a sample cell window.

1. Radially-invariant noise which has a constant offset at a fixed time, and has identical values at every radial position.

2. Normally distributed noise from non-linearity in the optics, intensity fluctuations of the lamp flashes, refractive artifacts, and systematic contributions of unknown sources.