1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
|
<!-- header fragment for html documentation -->
<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 3.2 Final//EN">
<HTML>
<HEAD>
<META NAME="description" CONTENT="Estimation of population parameters using genetic data usi
ng a maximum likelihood approach with Metropolis-Hastings Monte Carlo Markov chain importanc
e sampling">
<META NAME="keywords" CONTENT="MCMC, Markov chain, Monte Carlo, Metropolis-Hastings, populat
ion, parameters, migration rate, population size, recombination rate, maximum likelihood">
<TITLE>LAMARC Documentation: Changes</title>
</HEAD>
<BODY BGCOLOR="#FFFFFF">
<!-- coalescent, coalescence, Markov chain Monte Carlo simulation, migration
rate, effective population size, recombination rate, maximum likelihood -->
<P>(<A HREF="overview.html">Previous</A> | <A HREF="index.html">Contents</A>
| <A HREF="upcoming.html">Next</A>)</P>
<H2> Changes between LAMARC version 2.1.9 and 2.1.8 </h2>
<p><b>BUG FIX: corrected major divergence bugs.</b> Attempts
to infer epoch times produced markedly wrong results. All runs
inferring epoch times should be re-run.
</p>
<p><b>BUG FIX: spurious migration blocks in single population
converter runs.</b> The lam_conv file converter no longer creates
dysfunctional migration-enabled lamarc XML files when there is only
one population in the input data.
</p>
<H2> Changes between LAMARC version 2.1.8 and 2.1.6 </h2>
<p><b>BUG FIX: corrected major migration bug</b>
A bug in data summarization for migration parameters reversed the
direction of migration in some senses bug not others.
All runs modeling migration parameters not held constant
should be re-run.
</p>
<p><b>BUG FIX: improved handling of extreme negative growth</b>
A bug in tree generation under extreme negative growth resulted
in trees with default branch length being accepted.
If you have seen results (especially in Likelihood runs) where growth
rates oscillated regularly from negative to positive, that was
likely this bug.
All runs modeling growth should be re-run unless they were
Bayesian runs with exclusively non-negative priors.
</p>
<p><b>New feature: Divergence:</b>
LAMARC now models multiple populations
<a href="divergence.html">diverging from common ancestors</a>.
</p>
<p><b>New feature: SNP panel corrections:</b>
LAMARC can now correct for the loss of recent variation induced
by using <a href="panels.html">SNP panel chips</a>.
</p>
<p><b>Enhancement: faster convergence in Likelihood runs:</b>
Method for checking convergence in LIKELIHOOD runs has been improved,
yielding faster convergence.
</p>
<H2> Changes between LAMARC version 2.1.6 and 2.1.5 </h2>
<P><b>BUG FIX: corrected bug in data likelihood in presence of recombination</b>
A bug in the data likelihood calculation for recombinant trees caused
increasingly inaccurate data likelihoods as the number of non-recombinant
sub trees grew. This led to LAMARC preferring trees with fewer recombinations.
Recombinant analyses run with LAMARC versions 2.1.2 through 2.1.5 should
be re-run.
</P>
<P><b>BUG FIX: corrected random number generator</b>
Corrected a bug in which random values which were supposed to be in the
open interval (0,1) were sometimes returning 0 or 1. The bug was rare,
resulting in unexplained crashes every 4 billion or so random number draws.
There is no need to re-run analyses which did not crash.
We also updated our random number generators to use the
Boost Mersenne twister (boost::mt19937) random number generators.
</P>
<P><b>BUG FIX: data uncertainty model with SNP data (beta test)</b>
Data likelihood for the invariant base pairs in SNP data now
incorporates the per-base error rate.
Analyses using SNP data with the per-base error rate model should
be re-run.
</P>
<H2> Changes between LAMARC version 2.1.5 and 2.1.4 </h2>
<P><b>data uncertainty model (beta test)</b>
</P>
<P><b>improved output reports</b>
</P>
<H2> Changes between LAMARC version 2.1.4 and 2.1.3 </h2>
<P><b>compiles with g++ 4.3.3 on Linux</b>
This release updates the code base to compile with g++ 4.3.3.
Earlier compilers should still work.</P>
<P><b>minor user experience improvements</b>
Several error messages from the converter have been improved.
Converter can now read input files missing the end-of-line
character.
Lamarc now records the random seed used. This is useful
when debugging problems.
<H2> Changes between LAMARC version 2.1.3 and 2.1.2b </h2>
<P><b>Bug fix for haplotype rearranging code</b>
This release fixes a bug introduced into the code used to guess
haplotype resolution.
Analyses using this feature
in LAMARC versions 2.1.2 and 2.1.2b
should be re-run.
Additional improvements include:
<ul>
<li>Tracer output now renders step counts as integers instead of reals.
<li>Mapping output now writable to its own file.
<li>Upgraded default wxWidgets installation to 2.8.8.
</ul>
</P>
<H2> Changes between LAMARC version 2.1.2b and 2.1.2 </h2>
<P><b>Minor changes affecting user experience only</b>
Removed requirement for user to "press enter to quit" in
batch mode. Restored missing icons to MSW distribution.
</P>
<H2> Changes between LAMARC version 2.1.2 and 2.1.1 </h2>
<H3> Additions</H3>
<P><b>Limitations for Recombination relaxed.</b> LAMARC is now 'final
coalescent' aware, meaning that individual sites that have coalesced no
longer induce recombination events. This means that recombination can be
estimated for much longer distances--version 2.1.1 had problems when
theta * r * sequence length (or 4NCl) was any higher than about 5. 2.1.2
can now handle values of 4NCl up to ~100. In humans, this translates to
about .2 centimorgans, or 200 kilobases. We're working on expanding this
even further--if you have a data set that needs longer recombination
lengths, let us know (we're particularly interested in people using LAMARC
over long distances for trait mapping).</P>
<P><b>Tracer output</b> now includes the probabilities of the sequence data
on the current genealogy in Bayesian runs as well as in Likelihood runs.</P>
<h3>Corrections</h3>
<P><b>Unknown microsatellite data</b> are now recognized by the
converter.</P>
<P><b>Maximization</b> has been made somewhat more efficient in some cases,
particularly in runs that estimate variable mutation rates over regions
drawn from an unknown gamma distribution.</P>
<H2> Changes between LAMARC version 2.1.1 and 2.1. </h2>
<h3>Corrections</h3>
<P><b>Bayesian Tracer files</b> now omit parameters set to be 'invalid'.</P>
<P><b>Bayesian analyses</b> now handle runs with too few unique sampled
parameter values a bit more robustly, and warn a bit more sternly.</P>
<H2> Changes between LAMARC version 2.1 and 2.0.3. </H2>
<H3> Additions</H3>
<P><b>Trait mapping</b>. Trait data (such as disease status) can be mapped,
modelling trait changes as K-Allele data, with arbitrary models for
penetrance. Mapping can be performed using two approaches, one that
includes the trait data when rearranging trees, and one that analyzes the
trees after they are produced and collects the likelihoods of the data being
produced at each site. (See the <A HREF="mapping.html">mapping
documentation</a> for more information.)</P>
<P><b>Multiple data types within a linked genomic region.</b> LAMARC is now
able to correctly analyze a genomic region which contains, for example,
several microsatellite markers and a stretch of single-copy DNA. The
researcher will need to provide the expected relative mutation rate of each
type of data.</P>
<P><b>GUI Converter</b>. The GUI file conversion utility included with this
release has been significantly updated, replacing the beta version
originally released with version 2.0.</P>
<P><b>Batch versions</b>. LAMARC could previously be compiled in such a way
as to produce a 'batch' version. Now, that capability has been extended to
the normal version: If you execute lamarc with a '-b' (or '--batch')
command line option, it will run through and produce output without further
interaction from the user. The converter may be run the same way, also with
a '-b' command-line flag (see the <A HREF="converter.html">converter
documentation</a>).</P>
<P><b>Input file setting from the command line</b>. You may now specify a
LAMARC input file to use at the command line with a command like "lamarc
new_infile.xml". This is particularly helpful with the '-b' option, above,
as it means you can use a different input file than the default
'infile.xml'.</P>
<h3>Corrections</h3>
<P><b>Bayesian runs with different effective population sizes</b> for
different regions were producing erroneous output due to a bug in tree
rearrangement. (This would most easily manifest as abnormally low
acceptance rates for theta values in regions with an effective population
size other than 1.0.) This has been fixed.</P>
<h3> Incompatibilities</h3>
<P><b>Phase tags in XML input file.</B> As part of enabling multiple
data segments of different types within a linked genomic region, we
have changed the numbering system used to indicate which sites in an
individual are phase-unknown to match the numbering system used for
other purposes. As a result, any previous XML input files which explicitly
listed the phase-unknown sites for each individual (rather than
indicating that no sites were phase-known or phase-unknown) will need
to be hand-edited to use the <A HREF="xmlinput.html#phase">new scheme</a>.
LAMARC will generally be able to detect this problem and issue an error
message. If you find you have this problem with an old infile, you can
either recreate the lamarc input with the converter, or edit the infile.
If you have all phase-unknown data, the simplest method is to change the
default (and now incorrect) tags:</P>
<pre>
<phase type="unknown">
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 [...]
</phase>
</pre>
<P>To:</P>
<pre>
<phase type="known"> </phase>
</pre>
<P>The latter will set all of your markers to phase unknown.</P>
<br>
<H2> Changes between LAMARC version 2.0.2 and 2.0.3. </H2>
<P><em>Note: version 2.0.3 was not released to the general public, but was
made available for the 2006 <A
HREF="http://www.molecularevolution.org/">Workshop on Molecular
Evolution</a> at Woods Hole.</em></p>
<H3> Additions</H3>
<P> <B> Gamma rates among regions.</B> Mutation rates can now vary among
genomic regions according to a gamma distribution; the program will
attempt to estimate the gamma shape parameter. This is particularly
useful for collections of microsatellite regions where little is known
about the relative mutation rates of each region. See the document <A
HREF="gamma.html">"Combining data with different mutation rates"</a> for
more information.</P>
<P><B> Tracer compatibility.</B> LAMARC now automatically writes
files that can be read by the Tracer utility of Drummond and Rambaut.
In a Bayesian run Tracer can monitor convergence of the parameter
estimates. In a likelihood run it can monitor the data likelihood
of the genealogies. In both cases, it is useful in determining
whether the program has been run long enough. See <A
HREF="tracer.html">the "Using Tracer with LAMARC" documentation</a> for more information.</P>
<P><B> Newick tree.</B> LAMARC can now write out the tree of highest
data likelihood it finds for each region, as a Newick format tree, in cases
which do not have migration or recombination.</P>
<H3> Corrections </H3>
<P>Several limitations of the <b>Stepwise and Mixed-KS microsatellite
models</b> have been relaxed, allowing runs with widely divergent
microsatellite counts to run to completion instead of halting partway
through the program run. (If an analysis using an earlier version of LAMARC
ran to completion, it was not affected by this problem; affected runs
would crash.)</P>
<P>The mixing parameter of the <b>Mixed-KS model</b> could previously be set
to adjust during the run, but did not actually do so. Now it does.</P>
<br>
<H2> Changes between LAMARC version 2.0 and 2.0.2 </H2>
<H3> Additions</H3>
<P><B> Summary file reading and writing </B> can now be used with Bayesian
as well as likelihood analysis. </P>
<P><B> Maximizer fine-tuning </B> allows likelihood maximization and profiling
to succeed in some cases where they previously would have failed. </P>
<H3> Corrections </H3>
<P><B> Bayesian multi-region analysis was incorrect </B> in the previous version;
probability curves representing multiple regions were added rather than
multiplied. The resulting MPEs are correct but the confidence limits are
unnecessarily wide. All Bayesian runs with multiple unlinked genomic regions
should be redone. </P>
<P><B> Likelihood multi-replicate analysis was incorrect </B> in the previous
version. Neither MLEs nor confidence limits are accurate. All likelihood
runs done using replication from version 2.0 (not from earlier versions) should
be redone. </P>
<br>
<H2> Changes between LAMARC version 1.2 and 2.0 </H2>
<H3>Additions</H3>
<P><B> Bayesian analysis.</B> Lamarc can now make a Bayesian estimation of
population parameters as an alternative to the original maximum-likelihood
estimation. Linear and logarithmic priors with user-specified upper and
lower bounds are available. Users are strongly encouraged to set appropriate
priors. In our limited experience, the results of Bayesian analysis are
quite similar to those of likelihood analysis, but the Bayesian approach
may be superior for estimating parameter values near zero.</P>
<P>As an adjunct to Bayesian analysis, we offer a new genealogy-search
strategy which reconsiders only branch lengths. This may allow the
search to more rapidly react to newly proposed values of Theta. It can
also be used in likelihood-based analysis. It is currently enabled by
default for all lamarc runs, so attempts to exactly replicate previous
results will first need to disable this <A HREF="menu.html#rearrangers">strategy</A>.</P>
<P><B>Parameter constraints.</B> Individual population parameters (such
as migration or growth rates) may now be constrained to a user-specified
value, or groups of them may be constrained to be equal. We especially recommend
use of constraints to reduce the number of parameters in cases with
many subpopulations.</P>
<P><B>Different N<sub>e</sub> and mu among genetic regions.</B>
It is now possible to set the relative N<sub>e</sub> (effective
population size) and relative mu (neutral
mutation rate) of each genetic region independently, allowing a correct
joint estimate over unlike regions such as autosomal and sex-chromosome
samples or DNA and microsatellite samples.</P>
<P><B>New data types and models.</B> Data with multiple alleles among
which no particular relationship is implied, such as elecrophoretic
alleles, can now be coded as "K-Allele data" and analyzed via a K-Allele
model. This model is also available for microsatellites as an
alternative to the stepwise mutation models; furthermore, a mixed
model which attempts to optimize the ratio of stepwise changes and
K-allele changes can be used for microsatellite data.</P>
<P><B> Graphical user interface for file conversion utility.</B> File
conversion from PHYLIP or MIGRATE format files can now be done using
a GUI interface which is significantly easier than the text-based
form. The text-based converter is still available.</P>
<H3> Improvements and bug-fixes </H3>
<P><B>Multiple replicates</B> are now correctly implemented using the method of
Geyer.</P>
<P><B>Inference on phase-unknown DNA or SNP data</B> had a serious bug in version 1.2
which is fixed in this version. Previous analyses of this type should be
repeated as the bug did not cause a crash, but led to inaccurate results.</P>
<P><B> Effectiveness of the maximizer</B> has been greatly improved; it finds
correct maxima in a much larger proportion of cases. (The maximizer is the
part of the program that searches the n-dimensional likelihood surface for
the maximum height, which is the maximum likelihood.)</P>
<br>
<H2> Changes between LAMARC version 1.1 and 1.2 </H2>
<H3> Change in XML file format </H3>
<P> Older XML files which use the following tag will need to be modified. The
tag <map_position> (with an underscore) has become <map-position>
(with a dash) for consistency with the other tags. </P>
<H3> Removal</H3>
<P><B> Multiple replicates.</B> Regrettably, we have disabled the ability to
do multiple replicates of each chain as an accuracy improving measure. This
had not been implemented correctly; it produced approximately correct maximum
likelihood estimates but too-narrow confidence intervals. We will re-enable
this feature as soon as we have correct algorithms for combining results over
replicates. (This feature has been corrected and re-enabled in version 2.0.)</P>
<P>Previous multiple-replicate runs may well have too-narrow confidence intervals and
should be redone. We regret this problem. </P
<H3>Additions </H3>
<P><B> Growth.</B> The program can now estimate an exponential growth rate
for a single population or for several subpopulations. This duplicates the
functionality of FLUCTUATE except that LAMARC can estimate growth in the
presence of recombination and/or migration as well.</P>
<P><B>General Time-Reversible mutational model. </B> For DNA, RNA or
SNP data, the program can now use a fully-specified form of the GTR mutational
model. It is not able to optimize the parameters of this model, but
other tools such as PAUP*/Modeltest can be used to develop an optimal model
to be applied by LAMARC.</P>
<P><B>Adaptive heating.</B> When using the MC^3 or "heated chains"
strategy to improve searching, the program can now adjust the temperatures
of the heated chains automatically in an attempt to improve efficiency,
rather than relying on user-specified fixed temperatures.</P>
<P><B> Menu revision.</B> The menu has been extensively revised and now
has the capacity to undo multiple changes. In addition, a few options on
the menu have been moved to theoretically more reasonable spots. </P>
<P><B> Saving menu options.</B> The program automatically writes a
file, "menusettings_infile.xml", which contains the user's original infile updated
with the results of all changes made via the menu. This greatly
simplifies re-running a complicated case.</P>
<P><B> Saving sampled genealogies. </B> The program is now capable
of writing a file containing summaries of its sampled genealogies,
and can read that file back in and resume a run. This is useful
in recovering a run that has crashed, and can also be used to do
more complex analyses of the same genealogies. For example, you
may wish to do a quick run with no profiling in order to find the
best run parameters, and then re-analyze those genealogies with
profiling if they are satisfactory.</P>
<P><B>No-menu option.</B> The program can now be compiled in a no-menu
form which takes all of its input from the XML infile. This is
useful in designing large simulation studies and other batch runs. </P>
<P> <b>Output</b> The tables of data have been transposed so that what
used to be displayed in rows is now displayed in columns. This puts all
modified values of the same parameter in the same column, which should make
it easier to follow the changes.</P>
<H3> Corrections </H3>
<P> <B>File converter.</B> When input data was presented in Phylip
"interleaved" format it was truncated in the file converter. Also,
if multiple input sequences had the same Phylip-truncated name, the
converter would silently discard the duplicates.</P>
<P> <B> Maximizer accuracy </B> The maximization routines which generated
the maximum likelihood estimates, confidence limits and profiles for
population parameters were sometimes unsuccessful in finding the true
maxima, leading to incorrect estimates and inconsistent profiles.
While we cannot guarantee that the new routines will succeed in all
cases, they are greatly improved and also provide more feedback when
they fail.</P>
<br>
<H2> Changes between LAMARC version 1.0 and 1.1 </H2>
<H3>Additions </H3>
<P><B>Microsatellite data.</B> Both a stepwise mutational model
and a Brownian model are provided. Variable rates at different
microsatellite regions can be accommodated with a Felsenstein-Churchill
Hidden Markov model. Warning: the stepwise model is very slow,
and so has not been as thoroughly tested as the others.</P>
<P><B>SNP data.</B> We implement the "reconstituted DNA" model of
<A HREF="http://www.genetics.org/cgi/content/abstract/156/1/439">Kuhner et
al. 2000.</a> The user must provide map information showing the
location of the SNPs relative to each other in order to estimate
recombination rate; unmapped SNPs are usable for population size
and migration rate estimation only. </P>
<P><B>Genotypic data.</B> The program can now use data for which the
haplotypes are unknown. It searches among many different haplotype
resolutions. Be sure to use heating if you use this
option, as otherwise the search tends to become stuck. </P>
<H3> Corrections </H3>
<P><B>Speed.</B> The version 1.0 release still contained some debugging
code which slowed it down substantially (all likelihoods were
calculated twice). Version 1.1 should be quite a bit faster.</P>
<P><B>Lost data in converter.</B> The file conversion program silently lost
the last nucleotide of each sequence. This will have had a slight
effect on the results. If your sequences are very short you may
wish to re-run previous analyses.</P>
<P><B>Incorrect converter output.</B> Using the file converter for multiple
population cases produced defective LAMARC input files which could
not run successfully.</P>
<P><B>File converter flexibility.</B> The file converter is now able
to deal with a much wider array of input data.</P>
<P>(<A HREF="overview.html">Previous</A> | <A HREF="index.html">Contents</A>
| <A HREF="upcoming.html">Next</A>)</P>
<!--
//$Id: changes.html,v 1.53 2013/11/08 23:09:53 mkkuhner Exp $
-->
</BODY>
</HTML>
|
