For two examples, the eigenvalues obtained from `g_covar` are all zeros

User discussions
1

GROMACS version: 2024
GROMACS modification: Yes/No
Here post your question :

Dear all,

This time, I had run a pure protein MD simulation and wanted to try PCA analysis. However, the results turned out to be 0.

I used the GROMACS 2024-version (also tried the 2021-version). The unrestrained MD simulation only ran for 300 ps (in the order of energy minimization, restrained MD, and then regular unrestrained MD).

For the produced ‘md.xtc’ file, I used ‘trjconv’ with the “-pbc mol -center” to process the trajectory, and I selected the ‘protein’ group both two times.

Then, I used ‘trjconv’ again with “-fit rot+trans” to process the trajectory from the previous step, and I selected the ‘protein’ group both two times, resulting in the ‘aligned.xtc’ file.

Finally, I ran the command gmx covar -s md.tpr -f aligned.xtc -o eigenvalues.xvg -v eigenvectors.trr -xpma covar.xpm, and I selected the ‘alpha-C atoms’ both two times.

When I checked the eigenvalues file (eigenvalues.xvg), all the values were 0 (every frame).

Two weeks ago, I also tried a similar simulation with a protein-ligand complex (PDBID:3ATL), and the results were the same—all zeros.

Did I misunderstand something? I really appreciate your help and look forward to your replies.

Thank you.

2

It’s very strange. If I proceed the PCA analysis using the following steps, the results for pc1.xvg seem relatively normal (at least not the strange ’ all 0’). I ran: gmx anaeig -f mdfit.xtc -s md.tpr -v eigenvectors.trr -first 1 -proj pc1.xvg. Can I trust the results of this pc1.xvg （plz see the sceenshot）?
@ autoscale onread none
@ with g0
@ g0 on
@ title “projection on eigenvectors (nm)”
@ xaxis ticklabel off
@ world xmin 0
@ world xmax 300
@ world ymin -0.179389
@ world ymax 0.137194
@ view xmin 0.15
@ view xmax 0.85
@ view ymin 0.847667
@ view ymax 0.85
@ yaxis label “vec 1”
@ xaxis tick major 100
@ xaxis tick minor 50
@ xaxis ticklabel start type spec
@ xaxis ticklabel start -0
@ yaxis tick major 0.1
@ yaxis tick minor 0.05
@ yaxis ticklabel start type spec
@ yaxis ticklabel start -0.1
@ zeroxaxis bar on
@ zeroxaxis bar linestyle 3
0.0000 0.01570
1.0000 0.01164
2.0000 0.02112
3.0000 0.02698
4.0000 0.02144
5.0000 0.01375
6.0000 0.00992
7.0000 0.01068
8.0000 0.01953
9.0000 0.02149
10.0000 0.03067
11.0000 0.03182
12.0000 0.01499
13.0000 0.01117
14.0000 0.00930
15.0000 0.00357
16.0000 -0.00486
17.0000 -0.01421
18.0000 -0.01346
19.0000 -0.00188
20.0000 0.00046
21.0000 0.00270
22.0000 0.00287
23.0000 0.00995
24.0000 0.00531
25.0000 -0.00318
26.0000 -0.01000
27.0000 -0.00711
28.0000 0.00525
29.0000 0.02332
30.0000 0.03177
31.0000 0.03001
32.0000 0.03735
33.0000 0.03528
34.0000 0.03686
35.0000 0.04458
36.0000 0.04303
37.0000 0.05348
38.0000 0.06437
39.0000 0.06490
40.0000 0.06915
41.0000 0.07670
42.0000 0.07943
43.0000 0.08068
44.0000 0.08722
45.0000 0.07531
46.0000 0.06937
47.0000 0.04793
48.0000 0.05372
49.0000 0.06006
50.0000 0.06410
51.0000 0.05763
52.0000 0.04967
53.0000 0.05449
54.0000 0.05488
55.0000 0.06095
56.0000 0.06327
57.0000 0.06888
58.0000 0.07794
59.0000 0.07736
60.0000 0.07121
61.0000 0.07518
62.0000 0.06981
63.0000 0.05985
64.0000 0.06663
65.0000 0.06720
66.0000 0.07187
67.0000 0.08019
68.0000 0.08730
69.0000 0.08564
70.0000 0.08092
71.0000 0.06559
72.0000 0.08922
73.0000 0.10109
74.0000 0.10501
75.0000 0.10468
76.0000 0.10005
…

3

here is a sceenshot for the ‘all 0’ eigenvalues.xvg,

Command line:

gmx covar -s md.tpr -f mdfit.xtc -o eigenval.xvg -v eigenvec.trr -xpma covar.xpm

gmx covar is part of G R O M A C S:

GROwing Monsters And Cloning Shrimps

@ title “Eigenvalues of the covariance matrix”
@ xaxis label “Eigenvector index”
@ yaxis label “(nm\S2\N)”
@TYPE xy
1 0
2 0
3 0
4 0
5 0
6 0
7 0
8 0
9 0
10 0
11 0
12 0
13 0
14 0
15 0
16 0
17 0
18 0
19 0
20 0
21 0
22 0
23 0
24 0
25 0
26 0
27 0
28 0
29 0
30 0
31 0
32 0
33 0
…

4

I think you misunderstood PCA and the eigenvalues. The eigenvalues are not per frame but for each eigenvector. My guess is that you first only analyzed a single frame, which results in a covariance matrix with all zeros and thus the eigenvalues are all zero.