Molecular%20Weight%20Determination%20of%20Unknown%20Proteins%20for%20NASA/JPL%20PAIR%20Program%20%20August%2024,%202001

1
Molecular Weight Determination of Unknown
Proteinsfor NASA/JPL PAIR Program August 24,
2001

• Barbara Falkowski
• Falgun Patel
• Celia Smith

2
The Overall Goal

• To determine molecular weight of unknown
  electrophoresis data

3
Method to Achieve the Goal

• Measure distances of unknown standards with
  PhotoShop and Spotviewer
• Decide whether Spotviewer or Photoshop is the
  better measuring tool.

• Run models on standard proteins
• Decide which model(s) work the best for the
  standards
• Run model(s) on unknown proteins.

• Decide which model(s) worked the best on the
  unknowns

4
SpotViewer Disadvantages

• Did not measure dye-front distance
• One needed to go into Photoshop to mark or crop
  the dye-front distance.
• Spotviewer missed bands
• Did not always pick up bands that were thin,
  blurry or close together.
• Sometimes gave two measurement values to one band
• Or gave values that were associated with any
  band.
• Did not pick up very light bands.

5
PhotoShop Advantages

• Did not need assistance from another program.
• Not as time consuming
• Light bands could be more easily discerned
  through color inversion/manipulation of the
  image.
• This also worked well with tightly packed, thin
  and blurred bands.

6
Normal Image
Inverted/color Manipulated Image
7
Models Tested
Gels/Protein Used

• Vitelline Envelopes (VE) for two species
  (Strongylocentrotus purpuratus and Lytechinus
  pictus)
• Vitelline Envelopes for two methods (DTT and
  mechanically isolated)

• Quadratic Regression
• Quadratic Cross Validation
• SLIC
• Log-Linear Model
• Log-Log Model
• Local Linear Model
• Quadratic Interpolation

8
Which model worked the best?

• No single model was best for all of the gels.
• It was found that different models worked better
  for different gels.

Quadratic Regression Model - 15 Gel 1
S.purp/L.pictus VE DTT Removal
SLIC Model - Gradient Gel 2
Jelly Seminal Plasma VE Time
Courses
LOG-LOG Model - 12. 5 Gels
Gel 4 VE Tris Supernatant Time
Course and Gel 6 VE Tris Pellet Time Course
9
Why was the Quadratic Model chosen for the Gel 1?
10

• Took Quadratic Regression of standards to find
  the intercept and coefficients.
• Used the intercept and coefficients in the
  equation
• LOG MW RM2a RMb c

• Put the relative mobility of the unknowns into
  the equation to come up with the following
  results

11
Log Molecular Weight Results for 15 Gel
12
What type of Cross Validation was done?

• Quadratic Cross Validation using relative
  mobility and Log Molecular Weight
• Cross Validation was not chosen at all
• The predicted value for the missing band was not
  close the the actual value in any of the gel
  cases.

13
Results for Cross Validation Model on Standards
14
Why was the SLIC Model chosen for the Gradient
Gel 2 ?

• Residual Sum 0.00
• Residual Squared Sum 0.00
• Largest R2 0.99

15
Why was the SLIC Model was chosen for the Gel 2?
16
Compare Values

• SLIC Type Models
• Log( LN(MW) ) A B LN( -LN(RM) )
• Compare Log Molecular Weight
• X e ( LN( X ) )
• Convert Log( LN(MW) ) into Log( MW )
• Log( MW) Log( e LN(MW) )

17
Log Molecular Weight Results for SLIC
18
Graph result of SLIC Model
19
Why was the LOG-LOG Model Chosen for 12.5 Gels

• LOG-LOG Model worked best for the 12.5 Gels (Gel
  4 VE Tris Supernatant Time Course and Gel 6
  VE Tris Pellet Time Course)
• Small residuals
• R2 gt .9
• Residuals did not have large sections of positive
  or negative.

20
The Log-Log Model

• The Log-Log model is of the form
  Log(MW)abLog(RM)cLog(RM)2
• It incorporates the Log model and the quadratic
  model to make a more successful madel.

21
(No Transcript)
22
Predictions
23
(No Transcript)
24
Conclusion

• Different models worked better on different on
  certain gel types. The Quadratic Regression
  Model on the 15 gel, SLIC Model for the gradient
  gel and the LOG-LOG Model worked best for 12.
  gels. This process could be much improved if
  there was more data on the different gel types.

25
Thank You

• Open for Questions