comparison ctffind3 and ctffind4

Submitted by Leo on Fri, 2015-02-27 13:48

Forums

ctffind & ctftilt

Hi, I noticed some rather unusual results after running ctffind4 v.4.0.10 on images previously processed by ctffind3.
ctffind4 consistently gives much much smaller astigmatism and also sometimes defocus values are rounded, i.e. 36000 / 36000; or 27000 / 27000 (see below). What could be the reasons for this and which results are more reliable? Parameters as set in Relion were the same for both runs except low resolution for ctffind4 was 30 A instead of 100.

Here are some values from ctffind4 runs:

FoilHole_23849529_Data_23853395_23853396_20150124_0434_ctffind3.log: 16780.70 16787.48 24.37 0.14142 Final Values
FoilHole_23849530_Data_23853395_23853396_20150124_0436_ctffind3.log: 23918.29 23931.29 24.17 0.17659 Final Values
FoilHole_23849531_Data_23853395_23853396_20150124_0438_ctffind3.log: 29315.43 29305.33 -56.43 0.10756 Final Values
FoilHole_23849532_Data_23853395_23853396_20150124_0440_ctffind3.log: 17566.18 17578.99 27.16 0.21694 Final Values
FoilHole_23849533_Data_23853395_23853396_20150124_0442_ctffind3.log: 22265.44 22280.31 26.69 0.18815 Final Values
FoilHole_23849534_Data_23853395_23853396_20150124_0444_ctffind3.log: 27000.00 27000.00 -5.00 0.16794 Final Values
FoilHole_23849535_Data_23853395_23853396_20150124_0446_ctffind3.log: 14987.04 14998.98 22.08 0.20817 Final Values
FoilHole_23849536_Data_23853395_23853396_20150124_0447_ctffind3.log: 18959.78 18970.15 19.72 0.22278 Final Values
FoilHole_23849537_Data_23853395_23853396_20150124_0449_ctffind3.log: 23442.83 23459.68 19.17 0.20029 Final Values
FoilHole_23849538_Data_23853395_23853396_20150124_0451_ctffind3.log: 11520.95 11529.03 21.44 0.20332 Final Values
FoilHole_23849539_Data_23853395_23853396_20150124_0453_ctffind3.log: 17801.89 17790.73 -73.02 0.22321 Final Values
FoilHole_23849540_Data_23853395_23853396_20150124_0455_ctffind3.log: 22864.73 22854.96 -67.51 0.17824 Final Values
FoilHole_23849541_Data_23853395_23853396_20150124_0457_ctffind3.log: 10786.59 10781.91 -67.97 0.18405 Final Values
FoilHole_23849544_Data_23853395_23853396_20150124_0459_ctffind3.log: 17000.00 17008.26 28.00 0.20210 Final Values
FoilHole_23849545_Data_23853395_23853396_20150124_0501_ctffind3.log: 24604.90 24616.76 18.63 0.17367 Final Values
FoilHole_23849546_Data_23853395_23853396_20150124_0504_ctffind3.log: 15623.28 15631.65 24.60 0.18881 Final Values
FoilHole_23849547_Data_23853395_23853396_20150124_0506_ctffind3.log: 23177.45 23188.06 18.41 0.18205 Final Values
FoilHole_23849548_Data_23853395_23853396_20150124_0507_ctffind3.log: 30981.25 30998.92 19.29 0.16036 Final Values
FoilHole_23849549_Data_23853395_23853396_20150124_0509_ctffind3.log: 20012.26 20025.34 21.88 0.19460 Final Values
FoilHole_23849550_Data_23853395_23853396_20150124_0511_ctffind3.log: 25209.33 25194.94 -66.23 0.17545 Final Values
FoilHole_23849551_Data_23853395_23853396_20150124_0513_ctffind3.log: 29943.61 29960.88 23.22 0.16723 Final Values
FoilHole_23849553_Data_23853395_23853396_20150124_0515_ctffind3.log: 17375.82 17364.10 -65.64 0.20756 Final Values
FoilHole_23849554_Data_23853395_23853396_20150124_0517_ctffind3.log: 22500.00 22500.00 -2.50 0.17253 Final Values
FoilHole_23849555_Data_23853395_23853396_20150124_0519_ctffind3.log: 26433.17 26446.91 21.82 0.16720 Final Values
FoilHole_23849556_Data_23853395_23853396_20150124_0521_ctffind3.log: 14298.84 14289.44 -62.48 0.20120 Final Values
FoilHole_23849557_Data_23853395_23853396_20150124_0523_ctffind3.log: 18500.00 18500.00 -5.00 0.20984 Final Values
FoilHole_23849558_Data_23853395_23853396_20150124_0525_ctffind3.log: 22627.67 22642.31 22.40 0.20263 Final Values
FoilHole_23849559_Data_23853395_23853396_20150124_0527_ctffind3.log: 10375.82 10368.29 -56.57 0.18284 Final Values
FoilHole_23849560_Data_23853395_23853396_20150124_0530_ctffind3.log: 22627.91 22639.79 17.07 0.19369 Final Values
FoilHole_23849561_Data_23853395_23853396_20150124_0532_ctffind3.log: 29585.66 29603.29 16.90 0.18343 Final Values
FoilHole_23849562_Data_23853395_23853396_20150124_0533_ctffind3.log: 18314.02 18324.37 19.21 0.20235 Final Values
FoilHole_23849563_Data_23853395_23853396_20150124_0535_ctffind3.log: 23011.25 23000.00 -67.53 0.18447 Final Values
FoilHole_23849564_Data_23853395_23853396_20150124_0537_ctffind3.log: 27500.00 27500.00 -37.50 0.15858 Final Values
FoilHole_23849565_Data_23853395_23853396_20150124_0539_ctffind3.log: 15576.28 15570.02 -77.91 0.20702 Final Values
FoilHole_23849566_Data_23853395_23853396_20150124_0541_ctffind3.log: 20388.71 20399.20 17.15 0.18961 Final Values
FoilHole_23849567_Data_23853395_23853396_20150124_0544_ctffind3.log: 30102.82 30115.91 19.45 0.15667 Final Values
FoilHole_23849568_Data_23853395_23853396_20150124_0546_ctffind3.log: 20893.06 20880.40 -64.04 0.19157 Final Values
FoilHole_23849569_Data_23853395_23853396_20150124_0548_ctffind3.log: 28514.79 28500.42 -67.92 0.16314 Final Values
FoilHole_23849570_Data_23853395_23853396_20150124_0550_ctffind3.log: 36000.00 36000.00 -20.00 0.15458 Final Values
FoilHole_23849571_Data_23853395_23853396_20150124_0552_ctffind3.log: 27112.53 27123.04 14.77 0.15982 Final Values

Here are the same images after ctffind3:

FoilHole_23849529_Data_23853395_23853396_20150124_0434_ctffind3.log: 16380.01 17245.28 20.62 0.19714 Final Values
FoilHole_23849530_Data_23853395_23853396_20150124_0436_ctffind3.log: 23486.97 24345.62 18.41 0.22941 Final Values
FoilHole_23849531_Data_23853395_23853396_20150124_0438_ctffind3.log: 28876.58 29873.24 25.70 0.10870 Final Values
FoilHole_23849532_Data_23853395_23853396_20150124_0440_ctffind3.log: 17368.22 17907.52 20.36 0.25455 Final Values
FoilHole_23849533_Data_23853395_23853396_20150124_0442_ctffind3.log: 21952.87 22713.93 20.79 0.24272 Final Values
FoilHole_23849534_Data_23853395_23853396_20150124_0444_ctffind3.log: 26617.92 27609.15 19.86 0.23052 Final Values
FoilHole_23849535_Data_23853395_23853396_20150124_0446_ctffind3.log: 14794.08 15280.81 16.18 0.24930 Final Values
FoilHole_23849536_Data_23853395_23853396_20150124_0447_ctffind3.log: 18656.59 19301.19 20.00 0.28078 Final Values
FoilHole_23849537_Data_23853395_23853396_20150124_0449_ctffind3.log: 23107.47 23805.02 14.32 0.25714 Final Values
FoilHole_23849538_Data_23853395_23853396_20150124_0451_ctffind3.log: 11308.18 11680.44 15.13 0.25513 Final Values
FoilHole_23849539_Data_23853395_23853396_20150124_0453_ctffind3.log: 17588.04 18013.97 13.53 0.28891 Final Values
FoilHole_23849540_Data_23853395_23853396_20150124_0455_ctffind3.log: 22576.58 23133.77 16.17 0.23571 Final Values
FoilHole_23849541_Data_23853395_23853396_20150124_0457_ctffind3.log: 10668.26 10874.58 19.84 0.25042 Final Values
FoilHole_23849544_Data_23853395_23853396_20150124_0459_ctffind3.log: 16828.97 17298.26 15.58 0.25154 Final Values
FoilHole_23849545_Data_23853395_23853396_20150124_0501_ctffind3.log: 24290.03 25084.38 14.05 0.22829 Final Values
FoilHole_23849546_Data_23853395_23853396_20150124_0504_ctffind3.log: 15472.20 15881.19 16.64 0.23250 Final Values
FoilHole_23849547_Data_23853395_23853396_20150124_0506_ctffind3.log: 22878.02 23527.33 16.72 0.23386 Final Values
FoilHole_23849548_Data_23853395_23853396_20150124_0507_ctffind3.log: 30494.77 31456.29 15.70 0.21197 Final Values
FoilHole_23849549_Data_23853395_23853396_20150124_0509_ctffind3.log: 19727.16 20379.70 18.32 0.21964 Final Values
FoilHole_23849550_Data_23853395_23853396_20150124_0511_ctffind3.log: 24907.38 25695.29 16.63 0.21751 Final Values
FoilHole_23849551_Data_23853395_23853396_20150124_0513_ctffind3.log: 29559.37 30550.33 17.98 0.20977 Final Values
FoilHole_23849553_Data_23853395_23853396_20150124_0515_ctffind3.log: 17154.18 17673.38 18.35 0.26609 Final Values
FoilHole_23849554_Data_23853395_23853396_20150124_0517_ctffind3.log: 22225.27 22948.52 17.31 0.21222 Final Values
FoilHole_23849555_Data_23853395_23853396_20150124_0519_ctffind3.log: 26010.71 26932.01 22.35 0.22600 Final Values
FoilHole_23849556_Data_23853395_23853396_20150124_0521_ctffind3.log: 14111.00 14609.82 20.00 0.25363 Final Values
FoilHole_23849557_Data_23853395_23853396_20150124_0523_ctffind3.log: 18279.30 18689.61 19.87 0.26279 Final Values
FoilHole_23849558_Data_23853395_23853396_20150124_0525_ctffind3.log: 22316.96 23020.72 15.00 0.24193 Final Values
FoilHole_23849559_Data_23853395_23853396_20150124_0527_ctffind3.log: 10291.08 10527.07 26.85 0.24036 Final Values
FoilHole_23849560_Data_23853395_23853396_20150124_0530_ctffind3.log: 22345.62 23059.98 13.17 0.17415 Final Values
FoilHole_23849561_Data_23853395_23853396_20150124_0532_ctffind3.log: 29130.46 30125.07 12.47 0.22331 Final Values
FoilHole_23849562_Data_23853395_23853396_20150124_0533_ctffind3.log: 18063.91 18491.41 18.15 0.26139 Final Values
FoilHole_23849563_Data_23853395_23853396_20150124_0535_ctffind3.log: 22765.35 23317.44 17.11 0.24337 Final Values
FoilHole_23849564_Data_23853395_23853396_20150124_0537_ctffind3.log: 27110.93 27861.70 10.00 0.21982 Final Values
FoilHole_23849565_Data_23853395_23853396_20150124_0539_ctffind3.log: 15412.34 15725.41 10.24 0.24789 Final Values
FoilHole_23849566_Data_23853395_23853396_20150124_0541_ctffind3.log: 20178.51 20703.88 11.66 0.24076 Final Values
FoilHole_23849567_Data_23853395_23853396_20150124_0544_ctffind3.log: 29663.35 30580.38 14.12 0.19947 Final Values
FoilHole_23849568_Data_23853395_23853396_20150124_0546_ctffind3.log: 20559.04 21158.77 15.95 0.24684 Final Values
FoilHole_23849569_Data_23853395_23853396_20150124_0548_ctffind3.log: 28073.17 28858.09 18.45 0.22068 Final Values
FoilHole_23849570_Data_23853395_23853396_20150124_0550_ctffind3.log: 35469.04 36640.32 13.95 0.17215 Final Values
FoilHole_23849571_Data_23853395_23853396_20150124_0552_ctffind3.log: 26660.72 27624.94 15.73 0.22727 Final Values

Alexis Fri, 2015-02-27 13:55

Hi Leo,

I bet you have dAst = 0.0. This would explain both that ctffind4 finds lower astigmatism and that sometimes you get rounded values. I need to make this clear in the documentation, but as I explained in another thread in this forum, the meaning of dAst has changed in recent versions of ctffind4:
"The meaning of 0.0 for this value in ctffind3 was not well documented. It caused ctffind3 (and ctffind <= 4.0.8) to not impose any restraint on the level of astigmatism. This runs counter intuition I believe, since setting this to 0.0 might be expected to mean that we don't want to allow any/much astigmatism.
So I changed this. Now, if the user gives 0.0<=dAst<=10.0 (now called "expected/tolerated astigmatism" in ctffind4), there will be a very strong restraint on astigmatism. To switch off the restraint altogether, the user must set this value to something negative.
The default suggested by ctffind4 is 100.0A, which gives a soft restraint on astigmatism. I have already discussed this with Sjors and the next release of Relion will also use 100.0A as a default. I recommend you do this yourself in the meantime."

The reason this sometimes leads to rounded values is that the refinement step, which follows a brute-force search, sometimes gets nowhere because the restraint on astigmatism is so strong. In those cases, you should be getting a warning message from the minimiser.

I would recommend using the default value for dAst, of 100.0. I will amend the documentation now.

Thanks for your report!
Alexis

Alexis Fri, 2015-02-27 14:03

To avoid further confusion for Relion users, ctffind-4.0.13 will interpret dAst=0.0 the way ctffind3 did when running with --old-school-input.

Thanks again for the report.
Alexis

Leo Sun, 2015-03-01 12:35

Hi Alexis, thank you, I will try this.

Also, do I understand correctly that boxsize=512 given by Relion will be ignored by ctffind4 and the whole micrograph will be used?
Many thanks!
Leo

Alexis Sun, 2015-03-01 13:02

Hi Leo,

You are misunderstanding the box size parameter.

What happens in ctffind4 is:

  1. the Fourier transform of the micrograph is computed
  2. the amplitudes are kept
  3. this amplitude spectrum (a real image) is Fourier transformed, the transform is cropped to box_size (i.e. high freq terms are discarded), the cropped FT is back FT'ed. This down-sampled amplitude spectrum is what's used for fitting

In ctffind3:

  1. The micrograph is cut into a set of box_size*box_size patches
  2. The amplitude spectrum for each patch is computed (actually some patches are thrown away)
  3. The fitting is done on the average of the patch amplitude spectra

Hope this clarifies it
Alexis

Alexis Sun, 2015-03-01 17:36

To be honest, I don't know! I usually use 256 or 512. The smaller the box size, the less CPU time is needed. So you could go to 128 to save time. However, I think (though I can't say I've studied this enough to be confident) going to really small box sizes might lead to less accuracy in the fit.