E values for each sequence, if known... this should measure how
close the sequence is to the key sequence, and how likely the
match was in the database searched.
This function is just like the classifyMSF function that
marks up an entire profile, but this one can mark up only
a single sequence in a profile if you prefer.
All color scales return a color corresponding to a double between
0.0 and 1.0; if the double is out of bounds, the color corresponding
to the nearest endpoint of the scale should be returned.
If your prediction is yes/no, or if you are only interested in
stats on correctness and incorrectness, you can create a matrix
which is 2xN, where N is the number of possible categories
in reality, and data[0][i] indicates the number of wrong predictions
of things belonging to catogory i (you predicted them as something
other than i), while data[1][i] indicates the number of correct
predictions of things belonging to category i.
data contains the actual matrix itself; it's public so that
it can be manipulated directly (for speed, such as that is
in java) and so that you can do something like 'dm.data[0][0] = tmp'
as well as 'tmp = dm.data[0][0]' without needing 2 different
functions.
data contains the actual array itself; it's public so that
it can be manipulated directly (for speed, such as that is
in java) and so that you can do something like 'dv.data[0] = tmp'
as well as 'tmp = dv.data[0]' without needing 2 different
functions.
data contains the actual matrix itself; it's public so that
it can be manipulated directly (for speed, such as that is
in java) and so that you can do something like 'dm.data[0][0] = tmp'
as well as 'tmp = dm.data[0][0]' without needing 2 different
functions.
data contains the actual array itself; it's public so that
it can be manipulated directly (for speed, such as that is
in java) and so that you can do something like 'dv.data[0] = tmp'
as well as 'tmp = dv.data[0]' without needing 2 different
functions.
data contains the actual matrix itself; it's public so that
it can be manipulated directly (for speed, such as that is
in java) and so that you can do something like 'im.data[0][0] = tmp'
as well as 'tmp = im.data[0][0]' without needing 2 different
functions.
data contains the actual array itself; it's public so that
it can be manipulated directly (for speed, such as that is
in java) and so that you can do something like 'iv.data[0] = tmp'
as well as 'tmp = iv.data[0]' without needing 2 different
functions.
This draws a line of letters showing the actual 2ary
structure of a protein, from residue startRes
(expressed in the protein's internal numbering scheme;
0 = first residue, etc).
This draws a line of letters showing the actual 2ary
structure of a protein, from residue startRes to startRes+n
(expressed in the protein's internal numbering scheme;
0 = first residue, etc).
makes an emptry DVector object with a specified dimension
All data should be zeroed out... this is
not done explicitly, but relies on Java's default values.
Filter out distance gaps (inter-CA > 4 A or < 3.6 A),
returning a new set of proteins consisting of this protein
broken into fragments everywhere there is a problem.
makes an emptry FVector object with a specified dimension
All data should be zeroed out... this is
not done explicitly, but relies on Java's default values.
g_b matrix used in:
mclachlan, eisenberg - PNAS 84:4355 - 4358
Gribskov & Burgess
1986
Nucleic Acid Research
vol 14
p 6745 - 6763
this is mdm78, adjusted to have mean of 0, and sd of 1.
This guesses whether the current machine is little-endian,
and would therefore need to use the reverseEndian calls
when reading/writing DataStreams for use with C programs.
makes an emptry IVector object with a specified dimension
All data should be zeroed out... this is
not done explicitly, but relies on Java's default values.
This is basically a lame re-implementation of ScrollPane,
but it works with AWT 1.0 applets... can therefore use
for Netscape 4.0, which supports Java 1.1 with AWT 1.0.
The 'large' network set. 120 jurors (level 1 and 2 networks, 15
of which use reduced training)
trained on all training sets (0 through 14), 7 unreduced and 1
reduced juror per set.
Make from 2 superimposed proteins, using JDB's method.
1) Make table of everything in (fold) which is <= CUTOFF angstroms
from each residue in (seq).
2) Get longest consecutive stretch, and set them as aligned;
remove these residues from future consideration.
3) Repeat until nothing is left that has any matches.
This function is given a sequence and fold being aligned
in a global alignment, and a S+1 x F+1 matrix containing
the scores for aligning each monomer in the sequence to
each monomer in the fold.
This call allows you to resume an interrupted optimization
at a particular point; it allows the initial step size to
still be very small while keeping a large rho.
A scoring function that checks if the predicted 2ary structure
in the sequence matches the real secondary structure in the fold,
while weighting a match by the predicted frequency.
A scoring function return a identity matrix; this is
set up so that a sampling of all calls to score() over
a large set of sequences and folds will average 0.0
with a standard deviation of 1.0.
A scoring function that checks if the predicted 2ary structure
in the sequence matches the real secondary structure in the fold,
while weighting things by the predicted frequencies.
train, cutoff, predict for a single training
set, or group of them... returns actual
number of steps taken, avg error on cnet4, and
average error on cnet1's.