Trees and Alignments of Protein Phosphatases

The page archives the trees and the underlying alignments of protein phosphatases.

Human

We infer the trees of human protein phosphatases by fold and family.

Notes:

• Paladin, CC2, CC3, PHP, RTR1 hs three genes or less in human, so no tree is built for them.
• Click the cells to download the data sets.
• The alignments are in Phylip format; the trees are in Newick format.
• PhyML JTT and RAxML JTT are the trees inferred by JTT substitution matrix.
• PhyML best model and RAxML best model are the trees inferred by the best model given by Prottest, in the format of substitution matrix (e.g. LG, VT, JTT) + invariable sites (I) + Gamma (G) + amino acid frequency (F) (see PhyML or RAxML manuals).
  
• The DSP and PTEN families are close to each other in sequence and structure. Thus, we built the tree for the them together.

Nine genomes

In addition to human,we infer the trees of the protein phosphatases in nine organisms, including human, sea urchin, fruit fly, Caenorhabditis elegans, the sea anemone Nematostella vectensis, the sponge Amphimedon queenslandica, a unicellular choanoflagellate Monosiga brevicollis, Saccharomyces cerevisiae and Dictyostelium discoideum.

Methods

We aligned the phosphatase domains by PROMALS3D followed by manual adjustment. We inferred the tree by two methods: neighbour joining (NJ) and Maximum Likelihood (ML). We used MEGA (version 6.0) to infer NJ trees, and RAxML (version 8.2.9 and PhyML (version 3.0) to infer ML trees. We used JTT substitution model in both methods. We also use the best model from Prottest by Bayesian information criterion (BIC) to infer ML trees (see individual folds and families below).

For PhyML, we use the parameters:

-d aa -b 100 -m MATRIX -f e -v e -a e -s SPR --rand_start 5 --n_rand_starts 5 --r_seed 1234

where MATRIX is the name of the subtitution matrix, e.g. LG, JTT, WAG; "-f e" toggles on +F, "-v e" toggles on +I, and "-a e" toggles on +G.

For RAxML, we use the parameters:

-f a -m MODEL -x 1234 -p 5678 -# 100

where MODEL is the model, i.e. the substitution matrix + I + G + F (see RAxML's manual).

The best models chosen by difference measurements (AIC, AICc, BIC, DT) are basically similar, but the rankings are different. For example, below human PTP (the numbers are weight (ranking).

    model          AIC         AICc        BIC         DT
    LG+I+G+F       0.98(1)     0.00(3)     0.00(3)     0.01(4)     
    LG+I+G         0.02(2)     1.00(1)     1.00(1)     0.01(1)     
    LG+G+F         0.00(3)     0.00(4)     0.00(4)     0.01(3)     
    LG+G           0.00(4)     0.00(2)     0.00(2)     0.01(2) 

However, sometimes the best models chosen by different measurements are quite different. For example, human myotubularins:

    model          AIC         AICc        BIC         DT
    LG+G+F         0.74(1)     0.76(1)     0.00(5)     0.03(4)     
    LG+I+G+F       0.26(2)     0.24(2)     0.00(7)     0.03(3)     
    JTT+G+F        0.00(3)     0.00(3)     0.00(10)    0.01(26)    
    JTT+I+G+F      0.00(4)     0.00(4)     0.00(13)    0.01(25)    
    VT+G+F         0.00(5)     0.00(5)     0.00(12)    0.00(62)    
    VT+I+G+F       0.00(6)     0.00(6)     0.00(15)    0.00(59)    
    WAG+G+F        0.00(7)     0.00(7)     0.00(14)    0.01(36)    
    WAG+I+G+F      0.00(8)     0.00(9)     0.00(16)    0.01(35)    
    LG+G           0.00(9)     0.00(8)     0.97(1)     0.03(1)     
    LG+I+G         0.00(10)    0.00(10)    0.03(2)     0.03(2) 

And human DSP+PTEN:

    model          AIC         AICc        BIC         DT
    LG+G+F         0.73(1)     0.00(61)    0.00(3)     0.02(4)     
    LG+I+G+F       0.27(2)     0.00(62)    0.00(4)     0.02(3)     
    LG+G           0.00(3)     0.00(7)     0.93(1)     0.02(1)     
    LG+I+G         0.00(4)     0.00(22)    0.07(2)     0.02(2)