Skip to content

Tony Notes: Tree Differences

Tony E Lewis edited this page Sep 1, 2017 · 2 revisions

Tony Notes: Tree Differences

A record of investigations into differences between trees (because some of this analysis has previously had to be repeated).

1.10.150.120 (45 starting clusters) (NOW REMOVED FROM BENCHMARK SET)

Current dfx.pl

New code exactly reproduces the tree (sha1:133447c6308dceb88111671fccd3b6923a69717e) that the current dfx.pl obtains (look for 1.10.150.120.trace in 2017_05_10.gemma_recode) under these two modes:

  • simple_ordering.compass_wp_dummy_1st.windowed
  • tree_df_ordering.compass_wp_dummy_1st.windowed

Dave's

tree_df_ordering.compass_wp_dummy_2nd.windowed is very similar to Dave's but has 5.03e-38 for 1041 vs 3001 (1041 3001 3005 5.03e-38) whereas Dave's has 5.08e-38 (1041 3001 3005 5.08e-38).

The relevant comparison is between node 1041 and nodes 925, 975, and 1767:

925	975	2995	1.52e-42
[...]
1767	2995	3001	1.25e-42
[...]
1041	3001	3005	5.0?e-38

3.30.70.1470 (45 starting clusters)

Current dfx.pl and Dave's (which are identical)

tree_df_ordering.compass_wp_dummy_1st.windowed and tree_df_ordering.compass_wp_dummy_2nd.windowed are very similar to each other, the only difference being:

9c9
< 24    25      582     1.73e-43
---
> 24    25      582     1.76e-43

tree_df_ordering.compass_wp_dummy_2nd.windowed/tree.trace is the more similar to Dave's / the dfx.pl tree (which are identical), with diff being:

41,44c41,44
< 612   613     614     1.31e-23
< 610   614     615     3.15e-18
< 604   615     616     9.21e-13
< 560   616     617     6.84e-08
---
> 610   612     614     2.68e-13
> 613   614     615     1.22e-26
> 604   615     616     1.14e-13
> 560   616     617     1.92e-09

Here, the new code's 612 613 614 1.31e-23 makes more sense than 610 612 614 2.68e-13. Further, that better evalue of 1.31e-23 for 612 / 613 can be reproduced (in either direction) using the COMPASS models that dfx.pl built:

~/cath-gemma/Cath-Gemma/tools/compass/compass_db1Xdb2_241 -g 0.50001 -i /home/lewis/rsync_of_2017_09_01.gemma_rrrcde/projects/2017_09_01.gemma_rrrcde/3.30.70.1470/results/612.prof -j /home/lewis/rsync_of_2017_09_01.gemma_rrrcde/projects/2017_09_01.gemma_rrrcde/3.30.70.1470/results/613.prof -n 0
~/cath-gemma/Cath-Gemma/tools/compass/compass_db1Xdb2_241 -g 0.50001 -i /home/lewis/rsync_of_2017_09_01.gemma_rrrcde/projects/2017_09_01.gemma_rrrcde/3.30.70.1470/results/613.prof -j /home/lewis/rsync_of_2017_09_01.gemma_rrrcde/projects/2017_09_01.gemma_rrrcde/3.30.70.1470/results/612.prof -n 0

...and that beats the 2.68e-13 for 610 / 612 that, again, can be reproduced (in either direction) using the COMPASS models that dfx.pl built:

~/cath-gemma/Cath-Gemma/tools/compass/compass_db1Xdb2_241 -g 0.50001 -i /home/lewis/rsync_of_2017_09_01.gemma_rrrcde/projects/2017_09_01.gemma_rrrcde/3.30.70.1470/results/612.prof -j /home/lewis/rsync_of_2017_09_01.gemma_rrrcde/projects/2017_09_01.gemma_rrrcde/3.30.70.1470/results/610.prof -n 0
~/cath-gemma/Cath-Gemma/tools/compass/compass_db1Xdb2_241 -g 0.50001 -i /home/lewis/rsync_of_2017_09_01.gemma_rrrcde/projects/2017_09_01.gemma_rrrcde/3.30.70.1470/results/610.prof -j /home/lewis/rsync_of_2017_09_01.gemma_rrrcde/projects/2017_09_01.gemma_rrrcde/3.30.70.1470/results/612.prof -n 0

1.10.1390.10 (189 starting clusters)

tree_df_ordering.compass_wp_dummy_1st.windowed and tree_df_ordering.compass_wp_dummy_2nd.windowed are identical.

Current dfx.pl

Dave's

Differences between trees generated on legion / bchuckle

Apart from differences in the way 0 is printed, the differences for:

  • 1.10.1390.10
  • 3.20.20.120
  • 3.30.70.1470
  • 3.40.50.970

are:

$ diff bchuckle_trees/3.40.50.970/simple_ordering.compass_wp_dummy_1st.naive_lowest/tree.trace  legion_trees/3.40.50.970/simple_ordering.compass_wp_dummy_1st.naive_lowest/tree.trace
1247c1247
< 36485 36482   36486   6.55e-03
---
> 36485 36482   36486   4.50e-03
$ diff bchuckle_trees/3.40.50.970/simple_ordering.compass_wp_dummy_2nd.naive_lowest/tree.trace  legion_trees/3.40.50.970/simple_ordering.compass_wp_dummy_2nd.naive_lowest/tree.trace
1247c1247
< 36485 36482   36486   6.55e-03
---
> 36485 36482   36486   4.50e-03
$ diff bchuckle_trees/3.40.50.970/simple_ordering.mk_compass_db.naive_mean/tree.trace           legion_trees/3.40.50.970/simple_ordering.mk_compass_db.naive_mean/tree.trace
856c856
< 35658 36052   36095   7.98867268672031e-53
---
> 35658 36052   36095   7.76240311269971e-53
$ diff bchuckle_trees/3.40.50.970/tree_df_ordering.compass_wp_dummy_1st.naive_lowest/tree.trace legion_trees/3.40.50.970/tree_df_ordering.compass_wp_dummy_1st.naive_lowest/tree.trace
1247c1247
< 36485 36482   36486   6.55e-03
---
> 36485 36482   36486   4.50e-03
$ diff bchuckle_trees/3.40.50.970/tree_df_ordering.compass_wp_dummy_2nd.naive_lowest/tree.trace legion_trees/3.40.50.970/tree_df_ordering.compass_wp_dummy_2nd.naive_lowest/tree.trace
1247c1247
< 36485 36482   36486   6.55e-03
---
> 36485 36482   36486   4.50e-03
$ diff bchuckle_trees/3.40.50.970/tree_df_ordering.mk_compass_db.naive_mean/tree.trace          legion_trees/3.40.50.970/tree_df_ordering.mk_compass_db.naive_mean/tree.trace
856c856
< 35658 36052   36095   7.98867268672031e-53
---
> 35658 36052   36095   7.76240311269971e-53