-
DNA helix
Voor de fylogenie van schimmels wordt gebruik gemaakt van DNA sequenties
-
Electronica projecten
Een mooie hobby om zelf electronica projecten te solderen
-
Brazilië
Veel watervallen zoals Foz do Iguacú
-
Noorwegen
Fjorden
-
China
De Lange Muur
-
Vliegenzwam
Amanita muscaria - licht giftig
-
Histoplasma capsulatum
Dimorfe schimmel dat ernstige infecties kan veroorzaken
Introduction
This course is meant to work with different kinds of alignment programs and tree contruction software.
There is a lot of available software out there, free and paid, but also web portals to do the job.
Web portals, however, are often limited to the size of the data to prevent users uploading gigabytes of data.
Your computer, depending on the physics like cpu, available RAM etc. makes it possible to handle large data sets.
PhyloMain v6, that is available via this website, is using freely available software to do the job.
Much more software on this topic is available on the internet and this website.
Updated version: 2020-11-15
If you already downloaded the package then use the next download to update the PhyloMain program:
Create a folder .e.g. D:\PhyloMain. Move the zip file into this folder and unzip the file here. A folder PhyloMain is created that contains the starting program and the subfolders Examples, Manual and Progs.
Move the subfolder Progs to C:\Users\Public (in Dutch: C:\Gebruikers\Openbaar).
In D:\PhyloMain start the program by double-clicking PhyloMain.exe; there is a newer version available down at this webpage. It offers some benefits.
When everything is correct, you will see this window appearing on your monitor.
Comments and questions can be addressed via contact form.
Phylogeny
What is phylogeny? These are definitions found in dictionaries:
1. The evolutionary development and history of a species or trait of a species or of a higher taxonomic grouping of organisms
2. A model or diagram delineating such an evolutionary history
3. A similar model or diagram delineating the development of a cultural feature
4. the development or evolution of a particular group of organisms
5. the sequence of events involved in the evolutionary development of species or taxonomic group of organisms
Phonetics: faɪˈlɒdʒɪnɪ
A phylogeny is a hypothetical relationship between groups of organisms being compared. A phylogeny is often depicted using a phylogenetic tree, describing the evolutionary relationships between organisms.
It will take some days to fully explain and understand phylogeny, but there are some important terms you need to know if you want to read phylogenetic trees:
synapomorphy: trait or structure derived from a common ancestor
monophyly: group of organisms (clade) that share a most recent common ancestor
paraphyly: group of organisms that don't share a common ancestor
homology: trait or structure that was present in the common ancestor and persisted in the evolved lineages, but might differ in form or function
homoplasy: a shared trait or structure between two or more organisms, that did not evolve from a common ancestor, the opposite of homology
convergent evolution: process whereby species that are not closely related, independently evolve functionally or visually similar structures or traits
divergent evolution: response to abiotic or biotic factors whereby organisms may develop homologous traits or structures and leads to speciation
These are 5 common abiotic factors, that might influence speciation:
atmosphere, chemical elements, sunlight/temperature, wind and water, but also nutrient enrichment, nonliving things like rocks, mountains, air and soil.
Biotic factors are based on living things, like organic matter, ratio of predators and prey, population size, presence of phyto- and zooplankton etc.
To be prepared, it is advisable to visit this website:
https://biologydictionary.net/phylogeny/
Examples
In the installation directory (D:\PhyloMain) the 'Examples' folder contains a few example files:
ALL_Dermas.fas, Clades_align_nodup_wsl.fas and LSU33.fas.
During this course we will work with these example files.
These files are prepared for the course. If you use your own fasta files always be sure the file contains NO blank lines.
As you can see, we are going to use files in fasta format mainly. Fasta formatted files are the standard files nowadays.
There are still phylogenetic programs, that don't use fasta formatted files, like MrBayes, RaxML and PAUP. They use nexus formatted files.
Nexus formatted files consist of blocks of information that contain strain information, parameters, calculation setup etc.
The first line of every nexus file is #NEXUS.
Exercise
Do you remember, that we looked at D:\PhyloMain\Examples\Clades_align_nodup_wsl.fas ?
This file contains '>' lines with the piping character '|'.
Create a new fasta file with only the strain number and genus/species name for all Candida species. Try this before looking at the solution.
When you check the checkbox "Only '>' lines" and click 'Show' you will see non-Candida strains at the end of the list.
Why is this?
The search looks for any occurrence of 'Candida' in the '>' line. It seems, that these entries also have 'Candida', probably as a synonym.
Sequence Alignment
This is the start before creating any cladograms. There are a lot of alignment programs available for nucleotide sequences as well as amino-based sequences. But how reliable are these programs in aligning the dataset the right way. And what is the right way. You can imagine, that sequences that look very similar, most of the programs will not have any problem to align them. But what happens if the sequences are variable, even on species level. When you are dealing with alignments of 20 or 50 strains then it is still possible to edit the alignment manually. However, the amount of strains to do phylogeny becomes much larger as well as the length of the sequences. It becomes time consuming to edit 1000 strains with sequences of 3000 bp.
We have to rely on the quality of the alignment.
In PhyloMain 3 well-known alignment programs can be selected: Muscle, MAFFT and ClustalW. In general you could say that Muscle and ClustalW are used for somehow similar sequences, although they can handle sequences with higher variability. MAFFT has more options to choose from like local or global alignment. What program should I use? That is difficult to say. If you understand the contents of the dataset used, you might be able to use the right software. Otherwise it is a case of trial & error.
In PhyloMain select 'Start alignment'
Muscle is selected as default. Some parameters can be set, like iterations and -diags. For small datasets that contain similar sequences an number of iterations of 8 should suffice. Otherwise it should be set to 12 or 16. Muscle will stop iterating when it reached the best possible alignment. When sequences look very similar, e.g. there are several sequences of the same species with low variability checking -diags will cluster these sequences and treat them as one preventing unnecessary comparisons and finish the alignment faster.
MAFFT has more options. As default FFT-NS-i is selected. This protocol will try to find the best possible parameters for aligning based on the dataset. FFT-NS-x will create 1 or 2 trees and uses the trees as a reference to establish an improved alignment. The other options are based on the dataset and can be used if we know the contents of our dataset. The next schemes is trying to make an attempt to explain what are the differences.
global alignment (G-INS-i) - all residues can be aligned over the total length.
XXXXXXXXXXX-XXXXXXXXXXXXXXX
XX-XXXXXXXXXXXXXXX-XXXXXXXX
XXXXX----XXXXXXXX---XXXXXXX
XXXXX-XXXXXXXXXX----XXXXXXX
XXXXXXXXXXXXXXXX----XXXXXXX
local alignment (L-INS-i) - the dataset contains one alignable domain flanked by non-alignable residues.
ooooooooooooooooooooooooooooooooXXXXXXXXXXX-XXXXXXXXXXXXXXX------------------
--------------------------------XX-XXXXXXXXXXXXXXX-XXXXXXXXooooooooooo-------
------------------ooooooooooooooXXXXX----XXXXXXXX---XXXXXXXooooooooooo-------
--------ooooooooooooooooooooooooXXXXX-XXXXXXXXXX----XXXXXXXoooooooooooooooooo
--------------------------------XXXXXXXXXXXXXXXX----XXXXXXX------------------
affine gaps (E-INS-i) - the dataset contains alignable and non-alignable domains, where the non-alignable domains are not being aligned.
oooooooooXXX------XXXXooooooooooo---------------oooooooXXXXXooooooooooooooooo--ooooooooooooooo
---------XXXXX----XXXXoooooooooooooooooooooooooooooooooXXXXX-ooooooooooooooooooooooo----------
oooooooo-XXXXX----XXXX---------------------------------XXXXX---oooooooooo--oooooooooooo-------
---------XXXXXX---XXXX---------------------------------XXXXX----------------------------------
---------XXXXXXXXXXXXX---------------------------------XXXXX----------------------------------
---------XX-------XXXX---------------------------------XXXXX----------------------------------
'adjust direction' is checked. This will take care, that all sequences are in the same direction based on the first sequence in the file. If not, the sequence with a wrong direction will be reversed. The ID of the strain will be changed by adding '_R_' after the '>', e.g. >_R_CBS305.60
ClustalW uses iterations to establish a good alignment, mostly 3-5 iterations are used. It is possible to use a NJ reference tree with bootstrapping. It goes without saying, that calculation time will increase.
When looking at the two different aligment methods with MAFFT you will see that the alignment results are influenced by the method chosen. According to me the alignment with option L-INS-i looks better.
Let's have a closer look.
The sequence alignment now contains only lines with a strain number without Genbank numbers, e.g. >CBS_305.60 and without piping characters.
But the number doesn't say anything about the species we are dealing with. While 'Exporting selection' can reduce strain information 'Change strain label' can add information.
GBlocks
Gblocks eliminates poorly aligned positions and divergent regions of a DNA or protein alignment so that it becomes more suitable for phylogenetic analysis.
Gblocks can be started when an alignment is created by checking the GBlocks checkbox or use 'GBlock' in PhyloMain in an already created alignment.
Parameter values are difficult to guess; there are stringent and less stringent parameter values.
'Maximum Number of Contiguous Nonconserved Positions': the higher the number the more positions will be added to the final alignment (less stringent).
'Minimum Length of a Block': the higher the number the less positions will be added (more stringent).
'Allowed Gap Positions': there are three values, None, With half, All (how gaps are treated, more->less stringency).
None: all sites that contain gaps will be eliminated.
With half: sites that contain >50% gaps will be eliminated.
All: sites with gaps are allowed and treated as a separate value.
Unfortunately there is a downside to this program: GBlocks CAN NOT handle '>' lines larger than 50 characters.
Creating Trees
Finally, we created the best possible alignment. This aligned file can be used to create cladograms.
As with aligning, that has numerous programs to achieve this goal, there are also numerous programs available to create cladograms or phylogenetic trees.
There is difficulty in choice here too. What software should we use to obtain a tree with our dataset?
This also depends of the contents of our dataset. If the alignment contains very similar sequences it is no use to choose very complex tree construction programs.
It is like using a canon to kill a fly. It is time consuming and not very efficient. According to computer speed this is not really an issue anymore, because computers are relatively fast.
How to choose the right program? What do you want to achieve? A guide line to which program to use is shown in this picture.
Maximum parsimony: MEGA and PAUP are two software packages that are able to create these trees.
Distance methods: most tree constructing software is able to create trees based on distance matrices, MEGA, PAUP, FastTree, Phylip.
Maximum likelihood: these trees are made by RaxML, MrBayes, IQ-tree.
In PhyloMain we can use FastTree, IQ-tree, RaxML, PAUP and MrBayes to construct phylogenetic trees.
Creating maximum likelihood or parsimony trees is much slower than trees based on a distance matrix, like neighbor joining trees.
Distance methods are based on differences between the different sequences. All sequences are compared with each other resulting in a matrix.
| A | B | C | D | |
| A | - | 17 |
21 | 27 |
| B | - | 12 |
18 |
|
| C | - | 14 |
||
| D | - |
Based on these distances a tree can be created.
Maximum parsimony is trying to create the best possible tree out of all possible trees where the fewest required 'mutation events' are needed.
It uses only informative sites in an alignment, i.e. sites should have at least two or more different nucleotides and when there are only two then at least two or more sequences should have this nucleotide.
Seq1 ..AGCTAAAGGGTCAGGGGAAGGGCA..
Seq2 ..AGCAAAAGGGTCAGGGGAAGGGGA..
Seq3 ..AGCATAGGGGTCAGGGGAAAGGCT..
Seq4 ..AGCGGACCGGTAAGGAGAAAGGAC..
info ** * * **
With this method it is possible to obtain more than 1 tree that has the fewest 'mutation events'; which tree is the best tree can not be decided. In general a consensus tree is created with all 'best' trees with at least one polytomy, i.e. a tree with a branch that is not bifurcating. But an incompletely resolved tree is better than an incorrect tree.
Maximum likelihood is using all available sites and a lot of parameters based on a substitution model. Without proper substitution model a ML tree can not be created. Parameters, that are used, are e.g. tree and branch length, ratio of GC, GA, GT, AC, AT, TC, frequency of A, C, G, C etc.
Let's create a tree with the LSU33_muscle.afa alignment, that we created in the alignment course.
The alignment shows very similar sequences with conserved regions. We use 'FastTree', because IQ-tree, RaxML or MrBayes will not give a better tree and will take more computer time. FastTree was created to handle very large datasets.
FastTree
The tree only shows CBS numbers and no information about genus/species.
We are going to add information to the branches with data that is stored in an Excel file: LSU33_Data.xlsx
How to create such a datafile? Although it goes beyond the scope of this course, the following excercise uses BioEdit and Excel to create such a file relatively fast. I would assume, that you already have a list with strains you are working whether in Word or Excel.
In Excel you can manipulate data in many ways, but that is a totally different course.
IQ-tree
A more complex tree construction program is IQ-tree. This program has much more parameter settings and creates maximum likelihood trees. According to the flowchart already mentioned, this program is often used with a dataset that contains sequences with unrecognizable similarities.
Start IQ-tree by selecting 'Create IQ-tree' in PhyloMain.
This is the IQ-tree window:
Model criterion
IQ-tree creates maximum likelihood trees and therefor a substitution model is necessary. You can select your own substitution model by checking the box and choosing the model. Otherwise have the program calculate the best possible model based on the dataset; BIC (Bayesian Information Criterium) is the default algorithm and provides the easiest possible model. You can change this to another algorithm: AIC or AICc, Akaike Information Criterium or corrected Akaike Information Criterium.
IQ-tree parameters
Max. iterations doesn't have to be changed. IQ-tree seldomly uses all 1000 iterations to create a tree. The other options will establish values for the support of the branches in the tree.
The 'old' Nonparam BS (bootstrapping), that is also available in other programs, is slower than Bootstraps developed for IQ-tree (called UF-boot) and can't be used at the same time. Bootstrapping can be combined with aLRT iterations and Bayesian inference. When all three are used, the values will be shown in the tree. UF-boot should have a value >=1000, while non-parametric BS can have less iterations, 100 but preferably 500.
Note: when using Nonparametric BS it is assumed that values >70% are a good support for a branch. Due to the algorithm of UF-boot a well supported branch should have at least a value of 95%.
The use of a 'partition file' is only necessary in concatenated sequences of different markers. Different models of substitution can be assigned to the different markers in the alignment, for single marker alignment we don't have to create a partition file.
The rest of the parameters can be used to do extra tests or when a dataset contains a lot of shorter sequences. 'Prevent overestimating branch support' can be checked, if there is a possibility that the dataset will violate the conditions of the chosen model of substitution, i.e. the model parameters doesn't totally agree with the dataset, but there is no better model.
RaxML
RaxML can handle also very large datasets, but is very computer-intensive. The larger the dataset, the more power, memory and time is needed to build a phylogenetic tree.
Furthermore, RaxML is very sensitive when it comes to dataset formats. It use relaxed interleaved or sequential Phylip format or fasta format. The dataset has to be examined and edited to avoid the following problems:
1. Identical sequence name(s) appearing multiple times in an alignment, this can easily happen when you export a standard PHYLIP file from some tool which truncates the sequence names to 8 or 10 characters.
2. Identical sequence(s) that have different names, but are exactly identical. This mostly happens when you excluded some hard-to-align alignment regions from your alignment and does not make sense to use.
3. Undetermined column(s) that contain only ambiguous characters, that will be treated as missing data, i.e. columns that entirely consist of X, ?, *, for AA data and N, O, X, ?, for DNA data.
4. Undetermined sequence(s) that contain only ambiguous characters (see above) that will be treated as missing data.
Prohibited character(s) in taxon names are names, that contain any form of whitespace character, like blanks, tabulators, and carriage returns, as well as one of the following prohibited characters: colon, semi-colon, () or [].
RaxML can run parallel calculations (threads) depending on the amound of cores in your computer. How many threads you should use depends also on the size of the dataset, i.e. length of the alignment.
As a rule of thumb, one core/thread should be used for every 500 DNA sites. An alignment of 1000 sites will efficiently run on 2 threads, more threads might even decrease efficiency.
Bootstrapping
Bootstrapping is a commonly used procedure to get an insight in the strength/support of a clade. A value of 100(%) means the clade is strongly supported. Discussions about the minimum value of the bootstrap to decide whether a clade is a true clade is still going on. Some say it should be at least 80%, others agree on 70% or higher. Nevertheless, lower values doesn't always mean the clade is 'wrong'.
How does it work?
As an example we use this small alignment of 5 strains of 60 sites. The tree program will randomly choose 60 sites to create a new alignment set for the first bootstrap and creates a tree. Because the selection is random, some sites can be selected multiple times. This is repeated the amount of bootstraps that was set by the user, e.g. 1000. Mind you, this has nothing to do with the function of the gene or part of the gene, that is sequenced.
You can imagine, if the sequence of the strains are exactly the same, the order of nucleotides doesn't matter and will always create the same clade in these 1000 trees and will have a support value of 100. If however, the 'new' alignment causes some of the strains to be reallocated in a different clade the value will drop. At the end a consensus tree is created with the bootstrap values for each node.
PAUP
To create a parsimony bootstrap tree we use PAUP. This is a stand-alone program and will be started from within PhyloMain, but in a separate window.
PAUP is not using fasta-formatted sequence files, but in nexus-format. The conversion program in PhyloMain can convert sequence files.
The construction of the tree can be done in 3 ways: entering command lines in the PAUP window, via PAUP menus or add a PAUP-block at the end of the nexus file.
First we will start by using the command line facility in PAUP.
SequenceMatrix
This is a standalone program and will open in a separate window. It allows us to concatenate different markers to one sequence. It is programmed in Java and therefor a java runtime library should be installed on your computer.
Leading and trailing gaps will automatically be converted to ?
It is important, that the ID of strains ('>' line) in all sequence files are the same, otherwise the program can't concatenate them and will create a new entry.
>CBS 14054 is not the same as >CBS14054 or >CBS_14054 or >CBS 14054 Candida albicans.
During concatenation the program will inform us about the presence of a strain in the different sequence files. If not it will print (No data).
In the Examples directory there are no datasets for different genes, that can be concatenated. Extract 'sequencematrix.zip' in the Examples directory or download the following zip-file and extract it into the Examples directory. Three .afa files will be added to the directory, BT2.afa, EF1.afa and ITS.afa.
Watch this Youtube submission:
Similarity / Identity
MatGAT is a standalone program and will open in a separate window. It compares all sequences with each other and calculates the similarity and identity values. The table with values can be exported to Excel. Unfortunately, the colouring of the cells like shown in MatGAT is not exported. With a set of macro's in temp0.xlsm it is possible to add colours and some other formatting.
Select 'Similarity/Identity' in PhyloMain
How are similarity and identity calculated?
Suppose we have these 2 sequences aligned:
A: AAGGCTTCAGCTA
B: AAGGC--CTG-TA
Identity: nr. of identical nucleotides / shortest sequence length = 9/10 = 1 (90%)
Similarity: 1 - (nr. of differences / shortest sequence length) = 1 - (4/10) = 0.6 (60%)
