1 / 13

Background noise in STR testing

Background noise in STR testing. Travis Doom Dept. of Computer Science/Eng. Wright State University, Dayton, OH. Forensic Bioinformatics (www.bioforensics.com). Motivation. How is an electropherogram peak classified as a “true allele peak”, “technical artifact”, or “noise”?

betty_james
Download Presentation

Background noise in STR testing

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. Background noise in STR testing Travis Doom Dept. of Computer Science/Eng. Wright State University, Dayton, OH Forensic Bioinformatics (www.bioforensics.com)

  2. Motivation • How is an electropherogram peak classified as a “true allele peak”, “technical artifact”, or “noise”? • What are the sources of variation in lab RFU thresholds? • How might we determine these thresholds objectively?

  3. What is “Signal?” • Analytical methods rely on a measured signal for the analysis. • This signal is often a change in electronic voltage from an instrument. • The instrument measurement output (total/output/measurement signal) is the sum of the analyte signal and the nominal baseline/background signal. • St = Sx +Sb • The net signal indicates the concentration of the analyte. • Sx = St – Sb (the net signal) • Sx = gCx

  4. What is “Noise”? • For real instruments and samples, there are always some errors in determining baseline signal: • Drift: slow change in the baseline level over time • Noise: unwanted random fluctuations in the nominal baseline signal • Baseline = nominal +/- noise (~Sb) • St = Sx + ~Sb • If the baseline signal was exactly constant, any amount of analyte, no matter how small (within the resolution of the instrument) would be detectible. • Perfect correction for baseline could be made without error. • Perfectly precise results could always be obtained. • Noiseless experiments involving chemical concentrations and light detection instruments are not possible.

  5. How is analyte signal distinguished from noise? • Sx = St – ~Sb • Precision and accuracy of experimental results depend on: • The magnitude of the analyte signal • The magnitude of the noise • The limit to measuring/detection the presence of an analyte signal is not the size of the signal, nor the size of the noise alone, but the relative magnitude of the two. • Signal-to-Noise Ratio (S/N) • Noise can be characterized if reproducible • Assumption #1: Noise magnitude is independent of analyte signal magnitude • Assumption #2: Noise variance is Gaussian in nature • Thus, baseline signal is characterized by a mean μb and standard deviation σb • Role of the reagent blank

  6. Limit of Detection • Limit of detection: the lowest concentration of analyte [or analyte signal] that the analytical process can reliably detect (American Chemical Society). • The detection limit is based on a statistical calculation • Detection limit is closely related to the noise characterization of a series of measurements, including blanks under the same experimental conditions • Statistical certainty is expressed as a confidence limit for the noise error, generally 99.7% (3-SIGMA) • Sx = St - μb • Precisely correct only when Sb is μb. • As Sb varies due to noise, error varies into the reported net signal Sx. • Analyte “reliably detected” if: St – Sb >= 3 σb • Relative Magnitude (S/N): Sx >= 3 σb

  7. Thresholds for Detection/Quantization • How much output signal do we need to be reliably certain that we observe an analyte? • Detection threshold: μb + 3σb • Analyte signal at limit of detection = St - threshold • How much output signal do we need to be reliably certain that we can quantize the amount of analyte observed? • It is more difficult to determine the concentration of an analyte than merely to detect its presence or absence. • This is generally accepted as a 10-SIGMA decision • Quantization threshold: μb + 10σb

  8. Signal Measure Saturation Quantization limit μb + 10σb Measured signal (In Volts/RFUS/etc) μb + 3σb Detection limit Mean background Signal μb 0

  9. Objective threshold determination • The limit of detection is an extrapolated value. • While easy to use, carte blanche thresholds make assumptions that may not be valid for a particular experiment/run. • FBS study (currently unpublished) • Study characterizes noise signal in 42 runs taken from 7 cases analyzed by the FBI. • Each run contains a regent blank, a positive control, and a negative control. • Output signal data was collected only from regions of the electropherogram free of analyte signal (positive control peaks, ROX peaks, +/-4 stutter) in all channels. • In-line regent blanks/controls

  10. Study Results

  11. Positive Control (Average) Saturation 47 Quantization limit Measured signal (RFUs) 19 Detection limit Mean background Signal 4 0

  12. Negative Control (Maximum) Saturation 262 Quantization limit Measured signal (RFUs) 90 Detection limit Mean background Signal 25 0

  13. Contributions • Rubinson and Rubinson. Contemporary Instrumental Analysis. Prentice Hall, 2000. • Formal definitions in Section 5.5: Signals, Noise, and Detection limits • National Center for Biotechnology Information (NCBI) • BatchExtract • Forensic Bioinformatics • Jason Gilder

More Related