Secondary structure prediction of proteins in Jalview

Secondary structure prediction of proteins in Jalview


In this Jalview Online Training video, I will
consider how to do secondary structure prediction of proteins in Jalview. The prediction is performed using the Jpred
server at the University of Dundee. For more information, I recommend visiting
the JPred web site and reading their research papers. For this demonstration, I will import a group
of sequences that have already been aligned. The URL is shown on the screen, and is in
the notes below. JPred is run by going to the Web Services
drop down menu and selecting Secondary Structure Prediction,
JPred Secondary Structure Prediction. There are two ways that the JPred prediction
can be executed. JPred can make predictions from a single sequence
or from a pre-aligned set of sequences. When predicting from a single sequence, JPred
first uses PSIBLAST to search the Uniprot database and build a multiple
alignment. Predictions made after a PSIBLAST search are
usually more accurate but take longer to calculate. If a specific sequence is selected, when JPred
is launched, a full JPred prediction is executed. This involves a PSI-BLAST sequence database
search and an alignment profile generated to find similar
sequences. JPred then uses this PSI-BLAST alignment profile
to perform the secondary structure prediction. A status window opens reporting the progress
of the job. It can take a few minutes. Upon completion
a new alignment window opens, with the Jpred predictions results visible
in the annotations panel. Note that the number of sequences in the results
window is much larger than the original alignment. It is recommended to view the alignment, and
check that PSI-BLAST alignment is meaningful. If no sequence is selected, and sequences
are aligned, then JPred will not perform the PSI-BLAST step. Instead it uses the current alignment as the
profile to perform the secondary structure prediction. And JPred will use the first sequence in the
alignment for its prediction. Upon completion, a new alignment window opens. The number of sequences in this window matches
the number of sequences in the original alignment. Let us take a closer look at the JPred annotations. The consensus JPred prediction is displayed
in the row called JNetPRED. The helices are represented as red tubes,
and the strands are represented as green arrows. The confidence estimates for this prediction
is displayed in the JNetCONF row. Higher values mean higher confidence in the
accuracy of the prediction. The consensus JPred prediction is generated
from several outputs, such as the HMM and the PSSM profile based
predictions. Their results are shown on the JNetHMM and
JNETPSSM annotation rows respectively. The JNETJURY annotation row marks where all
the profile prediction outputs disagree. Lupas_21, Lupas_14, and Lupas_28 annotations
rows predict the presence of coiled-coils. The JNet burial annotation predict solvent
accessibility levels. Let us compare our two predictions. In the
second prediction window, I will highlight the sequence that was used
to generate the first Full Jpred prediction, that involved the PSI-BLAST step. Back in the first prediction window, I will
copying this sequence. By using Control or Command /C. And I will paste the sequences onto the alignment
window of the second prediction using Control or Command / V. The annotations associated with the sequence
have also been copied across. And we now have both predictions in a single
window. The sequence that I copied isn’t aligned
relative to this alignment. So I will do this by adding some gaps into
the sequence using the mouse and the Shift key. The annotation panel is looking quite busy. So I will hid some of the annotations rows. If you are not sure which annotation row
is associated with which prediction, then hovering the mouse over the annotation
brings up a tool tip with the name of the associated sequence. Now we can easily compare the results from
both predictions. Secondary structure prediction is very dependent
on the diversity of sequences and quality of the alignment. For this reason using PSI-BLAST to generate
the alignment profile, typically improve the secondary structure
prediction. The JPRED secondary structure prediction can
be performed on a subsection in an alignment, if is selection is aligned. JPRED uses this
selection as the profile to perform the secondary structure prediction. The first sequence in the set will be used
for a JPred prediction. If columns or part of the alignment is hidden. They will be exclude from the alignment that
is sent to the JPred service. The prediction results returned from the service will be mapped back onto the visible parts of the sequence, to ensure a single frame of reference is maintained. I can view all of the predictions in the original
alignment desktop window”. First I go to the original alignment, select
all the sequences, then right click the mouse to open the context
menu. Go to selection and select the Add Reference
Annotation option. All the JPred prediction results for the sequences
are now visible in the original alignment. The JPred prediction can be used to colour
the alignment by going to the colour menu, select colour by annotation and the select
the annotation required. If a 3D structure is known for one proteins
in the alignment, then JPred prediction can be used to colour
the structure. For more information, please look at our User
Manual on the the Jalview website at www.jalview.org. Good bye