| Author |
 Topic  |
|
|
bobmih
3 Posts |
 Posted - 11/09/2015 : 5:01:24 PM
|
Origin 2015:
Hello, I'm trying to figure out how to use Origin in the Resonant Recogniton Model (RRM). I've been trying to work my way through http://www.omicsonline.org/digital-signal-processing-techniques-calculating-biological-functionalities-jpb.1000199.pdf
Basically, a protein or DNA/RNA sequence is converted to a numeric value - each amino acid and nucleotide is assigned a specific electron ion interaction pseudopotential (EIIP) value.
These EIIP values are then transformed by FFT, STFT, DFT or wavelet to get a consenus frequency which is related to the activity of a protein or likelihood of a coding/binding sequence.
The above paper is the most basic I could find and I'm still having trouble with it :)
In the above linked paper there are two peptides (one of length 18 and one of length 32):
EIIP values of: (the p18 has been padded with 0’s)
P32 EIIP P18 EIIP
0 0.0058
0.0058 0.0548
0.0761 0.0829
0.0516 0.0198
0 0.0829
0.0371 0.0829
0.0371 0.0057
0 0.0941
0.0371 0.0829
0.1263 0.005
0.0829 0.1263
0 0.005
0.0829 0
0.0941 0.0761
0.0058 0.0057
0.0548 0.0959
0.0829 0
0.0198 0.0371
0.0829 0
0.0829 0
0.0057 0
0.0941 0
0.0829 0
0.005 0
0.1263 0
0.005 0
0 0
0.0761 0
0.0057 0
0.0959 0
0 0
0.0371 0
The text then states: Fast Fourier Transform (FFT): The numerical signals of Peptides P18 and P32 are then decomposed and processed by means of Fast Fourier Transform (FFT), a faster algorithm for the Discrete Fourier Transform (DFT) using equation:
y = fft(x)
The outcome of the Fourier Transform decomposition yields 32 sets of imaginary, real and absolute values called Spectral Characteristics (SC), which graphically describes the bio-recognition and bio-adhesion of the protein residues/peptides. The plot of SC presents a symmetric (mirror) image [29].
As a result, half the values are considered and the x-axis is scaled to 0.5. The zero frequency of the spectrum also termed” DC component” which is the average value of the signal is discarded.
Consequently, 15 sets of values are obtained from the decomposition of the 32 protein residues.
The Spectral Characteristics (SC) and Cross Spectral (CS) features of the P18 and P32 are shown below
Spectral Spectral
Features Features CrossSpectral
(Values) (Values)
p18 p32 P18 and p32
0.4778 0.632 0.30197
0.621 0.401 0.249021
0.583 0.182 0.106106
0.425 0.646 0.27455
0.5025 0.396 0.19899
0.617 0.389 0.240013
0.176 0.266 0.046816
0.09 0.399 0.03591
0.597 0.28 0.16716
0.697 0.797 0.555509
0.738 0.55 0.4059
1 1 1
0.441 0.35 0.15435
0.538 0.377 0.202826
0.585 0.755 0.441675
The Cross-Spectral analysis represents the point-wise multiplication of the Spectral Characteristics. The length of the largest peptide P32 is 32. There are no protein residues at positions 19 to 32 of the P18 and as such, they are zero-padded. The sequences are then translated using EIIP, processed by means DFT to yield Spectral Characteristics (SC) and point-wise multiplied to generate the Cross Spectral (CS) features.
What I've tried to do is use the FFT function under analysis>signal processing and clicked the different windowing and normalization options to try and get numbers close to the spectral values above. This has been to no avail.
Does anyone have any ideas on how to duplicate their results?
Any help would be appreciated.
Thank you,
Bob |
|
|
AmandaLu
439 Posts |
Posted - 11/10/2015 : 05:19:13 AM
|
Hi Bob,
I am not sure what software the reference used to perform FFT, but the results seem not quite right… FFT yields 32 sets. The first one (frequency = 0) is not included, and the left 32 sets should be mirror symmetric: 2nd – 16th vs. 18th-32nd, just as results yield by Origin. Window won’t change the symmetry. However, the results in reference show the symmetry is 1st-16th vs. 17th-32nd. We also compare the results with other software and it shows our results are correct.
Thanks,
Amanda
OriginLab Technical Service |
 |
|
|
bobmih
3 Posts |
Posted - 11/10/2015 : 09:27:36 AM
|
Thank you,
I think the authors of the paper may be leaving a step or two out- perhaps a normalization and something else.
|
|
|
|
AmandaLu
439 Posts |
Posted - 11/10/2015 : 8:53:25 PM
|
Hi Bob,
Neither normalization nor something change the symmetry. I supposed the author used some very old method to calculate FFT. The current method works out just like Origin.
Thanks,
Amanda
OriginLab Technical Service |
|
|
|
bobmih
3 Posts |
Posted - 11/13/2015 : 4:24:29 PM
|
Thanks everyone,
The problem as they say in coding was between the chair and keyboard.
The authors used a Rectangular window and having read somewhere else about the Hamming window (and not knowing the difference) was applying the Hamming window.
The authors then ran the analysis on each sequence and were calling what in Origin is the Magnitude column - the spectral power density.
They then multiplied the magnitudes together and then normalized them to 1.
|
|
|
| |
Topic |
|
|
|
help with FFT application