static void singleBootstrap(tree *tr, int i, analdef *adef, rawdata *rdta, cruncheddata *cdta)
{
tr->treeID = i;
tr->checkPointCounter = 0;
computeNextReplicate(tr, &adef->boot, (int*)NULL, (int*)NULL, FALSE, FALSE);
initModel(tr, rdta, cdta, adef);
getStartingTree(tr, adef);
computeBIGRAPID(tr, adef, TRUE);
if(adef->bootstrapBranchLengths)
{
switch(tr->rateHetModel)
{
case GAMMA:
case GAMMA_I:
modOpt(tr, adef, TRUE, adef->likelihoodEpsilon);
break;
case CAT:
tr->likelihood = unlikely;
catToGamma(tr, adef);
initModel(tr, rdta, cdta, adef);
if(i == 0)
modOpt(tr, adef, TRUE, adef->likelihoodEpsilon);
else
modOpt(tr, adef, TRUE, adef->likelihoodEpsilon);
gammaToCat(tr);
break;
default:
assert(0);
}
}
printBootstrapResult(tr, adef, TRUE);
}
void shSupports(tree *tr, analdef *adef, rawdata *rdta, cruncheddata *cdta)
{
double
diff,
*lhVectors[3];
char
bestTreeFileName[1024],
shSupportFileName[1024];
FILE
*f;
int
interchanges = 0,
counter = 0;
assert(adef->restart);
tr->resample = permutationSH(tr, 1000, 12345);
lhVectors[0] = (double *)rax_malloc(sizeof(double) * tr->cdta->endsite);
lhVectors[1] = (double *)rax_malloc(sizeof(double) * tr->cdta->endsite);
lhVectors[2] = (double *)rax_malloc(sizeof(double) * tr->cdta->endsite);
tr->bInf = (branchInfo*)rax_malloc(sizeof(branchInfo) * (tr->mxtips - 3));
initModel(tr, rdta, cdta, adef);
getStartingTree(tr, adef);
if(adef->useBinaryModelFile)
{
readBinaryModel(tr);
evaluateGenericInitrav(tr, tr->start);
treeEvaluate(tr, 2);
}
else
modOpt(tr, adef, FALSE, 10.0);
printBothOpen("Time after model optimization: %f\n", gettime() - masterTime);
printBothOpen("Initial Likelihood %f\n\n", tr->likelihood);
do
{
double
lh1,
lh2;
lh1 = tr->likelihood;
interchanges = encapsulateNNIs(tr, lhVectors, FALSE);
evaluateGeneric(tr, tr->start);
lh2 = tr->likelihood;
diff = ABS(lh1 - lh2);
printBothOpen("NNI interchanges %d Likelihood %f\n", interchanges, tr->likelihood);
}
while(diff > 0.01);
printBothOpen("\nFinal Likelihood of NNI-optimized tree: %f\n\n", tr->likelihood);
setupBranchInfo(tr->start->back, tr, &counter);
assert(counter == tr->mxtips - 3);
interchanges = encapsulateNNIs(tr, lhVectors, TRUE);
strcpy(bestTreeFileName, workdir);
strcat(bestTreeFileName, "RAxML_fastTree.");
strcat(bestTreeFileName, run_id);
Tree2String(tr->tree_string, tr, tr->start->back, FALSE, TRUE, FALSE, FALSE, FALSE, adef, SUMMARIZE_LH, FALSE, FALSE);
f = myfopen(bestTreeFileName, "wb");
fprintf(f, "%s", tr->tree_string);
fclose(f);
strcpy(shSupportFileName, workdir);
strcat(shSupportFileName, "RAxML_fastTreeSH_Support.");
strcat(shSupportFileName, run_id);
Tree2String(tr->tree_string, tr, tr->start->back, TRUE, TRUE, FALSE, FALSE, FALSE, adef, SUMMARIZE_LH, FALSE, TRUE);
f = myfopen(shSupportFileName, "wb");
fprintf(f, "%s", tr->tree_string);
fclose(f);
printBothOpen("RAxML NNI-optimized tree written to file: %s\n", bestTreeFileName);
printBothOpen("Same tree with SH-like supports written to file: %s\n", shSupportFileName);
printBothOpen("Total execution time: %f\n", gettime() - masterTime);
exit(0);
}
void doInference(tree *tr, analdef *adef, rawdata *rdta, cruncheddata *cdta)
{
int i, n;
#ifdef _WAYNE_MPI
int
j,
bestProcess;
#endif
double loopTime;
topolRELL_LIST *rl = (topolRELL_LIST *)NULL;
int
best = -1,
newBest = -1;
double
bestLH = unlikely;
FILE *f;
char bestTreeFileName[1024];
double overallTime;
n = adef->multipleRuns;
#ifdef _WAYNE_MPI
if(n % processes != 0)
n = processes * ((n / processes) + 1);
#endif
if(!tr->catOnly)
{
rl = (topolRELL_LIST *)rax_malloc(sizeof(topolRELL_LIST));
initTL(rl, tr, n);
}
#ifdef _WAYNE_MPI
long parsimonySeed0 = adef->parsimonySeed;
n = n / processes;
#endif
if(adef->rellBootstrap)
{
#ifdef _WAYNE_MPI
tr->resample = permutationSH(tr, NUM_RELL_BOOTSTRAPS, parsimonySeed0 + 10000 * processID);
#else
tr->resample = permutationSH(tr, NUM_RELL_BOOTSTRAPS, adef->parsimonySeed);
#endif
tr->rellTrees = (treeList *)rax_malloc(sizeof(treeList));
initTreeList(tr->rellTrees, tr, NUM_RELL_BOOTSTRAPS);
}
else
{
tr->resample = (int *)NULL;
tr->rellTrees = (treeList *)NULL;
}
for(i = 0; i < n; i++)
{
#ifdef _WAYNE_MPI
if(i == 0)
{
if(parsimonySeed0 != 0)
adef->parsimonySeed = parsimonySeed0 + 10000 * processID;
}
j = i + n * processID;
tr->treeID = j;
#else
tr->treeID = i;
#endif
tr->checkPointCounter = 0;
loopTime = gettime();
initModel(tr, rdta, cdta, adef);
if(i == 0)
printBaseFrequencies(tr);
getStartingTree(tr, adef);
computeBIGRAPID(tr, adef, TRUE);
#ifdef _WAYNE_MPI
if(tr->likelihood > bestLH)
{
best = j;
bestLH = tr->likelihood;
}
if(!tr->catOnly)
saveTL(rl, tr, j);
#else
if(tr->likelihood > bestLH)
{
best = i;
bestLH = tr->likelihood;
}
if(!tr->catOnly)
saveTL(rl, tr, i);
#endif
loopTime = gettime() - loopTime;
writeInfoFile(adef, tr, loopTime);
}
assert(best >= 0);
#ifdef _WAYNE_MPI
MPI_Barrier(MPI_COMM_WORLD);
n = n * processes;
#endif
if(tr->catOnly)
{
printBothOpenMPI("\n\nNOT conducting any final model optimizations on all %d trees under CAT-based model ....\n", n);
printBothOpenMPI("\nREMEMBER that CAT-based likelihood scores are meaningless!\n\n", n);
#ifdef _WAYNE_MPI
if(processID != 0)
{
MPI_Finalize();
exit(0);
}
#endif
}
else
{
printBothOpenMPI("\n\nConducting final model optimizations on all %d trees under GAMMA-based models ....\n\n", n);
#ifdef _WAYNE_MPI
n = n / processes;
#endif
if(tr->rateHetModel == GAMMA || tr->rateHetModel == GAMMA_I)
{
restoreTL(rl, tr, best);
evaluateGenericInitrav(tr, tr->start);
if(!adef->useBinaryModelFile)
modOpt(tr, adef, FALSE, adef->likelihoodEpsilon);
else
{
readBinaryModel(tr, adef);
evaluateGenericInitrav(tr, tr->start);
treeEvaluate(tr, 2);
}
bestLH = tr->likelihood;
tr->likelihoods[best] = tr->likelihood;
saveTL(rl, tr, best);
tr->treeID = best;
printResult(tr, adef, TRUE);
newBest = best;
for(i = 0; i < n; i++)
{
#ifdef _WAYNE_MPI
j = i + n * processID;
if(j != best)
{
restoreTL(rl, tr, j);
evaluateGenericInitrav(tr, tr->start);
treeEvaluate(tr, 1);
tr->likelihoods[j] = tr->likelihood;
if(tr->likelihood > bestLH)
{
newBest = j;
bestLH = tr->likelihood;
saveTL(rl, tr, j);
}
tr->treeID = j;
printResult(tr, adef, TRUE);
}
if(n == 1 && processes == 1)
printBothOpen("Inference[%d] final GAMMA-based Likelihood: %f tree written to file %s\n", i, tr->likelihoods[i], resultFileName);
else
printBothOpen("Inference[%d] final GAMMA-based Likelihood: %f tree written to file %s.RUN.%d\n", j, tr->likelihoods[j], resultFileName, j);
#else
if(i != best)
{
restoreTL(rl, tr, i);
evaluateGenericInitrav(tr, tr->start);
treeEvaluate(tr, 1);
tr->likelihoods[i] = tr->likelihood;
if(tr->likelihood > bestLH)
{
newBest = i;
bestLH = tr->likelihood;
saveTL(rl, tr, i);
}
tr->treeID = i;
printResult(tr, adef, TRUE);
}
if(n == 1)
printBothOpen("Inference[%d] final GAMMA-based Likelihood: %f tree written to file %s\n", i, tr->likelihoods[i], resultFileName);
else
printBothOpen("Inference[%d] final GAMMA-based Likelihood: %f tree written to file %s.RUN.%d\n", i, tr->likelihoods[i], resultFileName, i);
#endif
}
}
else
{
catToGamma(tr, adef);
#ifdef _WAYNE_MPI
for(i = 0; i < n; i++)
{
j = i + n*processID;
rl->t[j]->likelihood = unlikely;
}
#else
for(i = 0; i < n; i++)
rl->t[i]->likelihood = unlikely;
#endif
initModel(tr, rdta, cdta, adef);
restoreTL(rl, tr, best);
resetBranches(tr);
evaluateGenericInitrav(tr, tr->start);
modOpt(tr, adef, TRUE, adef->likelihoodEpsilon);
tr->likelihoods[best] = tr->likelihood;
bestLH = tr->likelihood;
saveTL(rl, tr, best);
tr->treeID = best;
printResult(tr, adef, TRUE);
newBest = best;
for(i = 0; i < n; i++)
{
#ifdef _WAYNE_MPI
j = i + n*processID;
if(j != best)
{
restoreTL(rl, tr, j);
resetBranches(tr);
evaluateGenericInitrav(tr, tr->start);
treeEvaluate(tr, 2);
tr->likelihoods[j] = tr->likelihood;
if(tr->likelihood > bestLH)
{
newBest = j;
bestLH = tr->likelihood;
saveTL(rl, tr, j);
}
tr->treeID = j;
printResult(tr, adef, TRUE);
}
if(n == 1 && processes == 1)
printBothOpen("Inference[%d] final GAMMA-based Likelihood: %f tree written to file %s\n", i, tr->likelihoods[i], resultFileName);
else
printBothOpen("Inference[%d] final GAMMA-based Likelihood: %f tree written to file %s.RUN.%d\n", j, tr->likelihoods[j], resultFileName, j);
#else
if(i != best)
{
restoreTL(rl, tr, i);
resetBranches(tr);
evaluateGenericInitrav(tr, tr->start);
treeEvaluate(tr, 2);
tr->likelihoods[i] = tr->likelihood;
if(tr->likelihood > bestLH)
{
newBest = i;
bestLH = tr->likelihood;
saveTL(rl, tr, i);
}
tr->treeID = i;
printResult(tr, adef, TRUE);
}
if(n == 1)
printBothOpen("Inference[%d] final GAMMA-based Likelihood: %f tree written to file %s\n", i, tr->likelihoods[i], resultFileName);
else
printBothOpen("Inference[%d] final GAMMA-based Likelihood: %f tree written to file %s.RUN.%d\n", i, tr->likelihoods[i], resultFileName, i);
#endif
}
}
assert(newBest >= 0);
#ifdef _WAYNE_MPI
if(processes > 1)
{
double
*buffer = (double *)rax_malloc(sizeof(double) * processes);
for(i = 0; i < processes; i++)
buffer[i] = unlikely;
buffer[processID] = bestLH;
for(i = 0; i < processes; i++)
MPI_Bcast(&buffer[i], 1, MPI_DOUBLE, i, MPI_COMM_WORLD);
bestLH = buffer[0];
bestProcess = 0;
for(i = 1; i < processes; i++)
if(buffer[i] > bestLH)
{
bestLH = buffer[i];
bestProcess = i;
}
rax_free(buffer);
}
if(processID == bestProcess)
{
#endif
restoreTL(rl, tr, newBest);
evaluateGenericInitrav(tr, tr->start);
printBothOpen("\n\nStarting final GAMMA-based thorough Optimization on tree %d likelihood %f .... \n\n", newBest, tr->likelihoods[newBest]);
Thorough = 1;
tr->doCutoff = FALSE;
treeOptimizeThorough(tr, 1, 10);
evaluateGenericInitrav(tr, tr->start);
printBothOpen("Final GAMMA-based Score of best tree %f\n\n", tr->likelihood);
strcpy(bestTreeFileName, workdir);
strcat(bestTreeFileName, "RAxML_bestTree.");
strcat(bestTreeFileName, run_id);
Tree2String(tr->tree_string, tr, tr->start->back, TRUE, TRUE, FALSE, FALSE, TRUE, adef, SUMMARIZE_LH, FALSE, FALSE, FALSE, FALSE);
f = myfopen(bestTreeFileName, "wb");
fprintf(f, "%s", tr->tree_string);
fclose(f);
if(adef->perGeneBranchLengths)
printTreePerGene(tr, adef, bestTreeFileName, "w");
#ifdef _WAYNE_MPI
}
#endif
}
if(adef->rellBootstrap)
{
//WARNING the functions below need to be invoked after all other trees have been printed
//don't move this part of the code further up!
int
i;
#ifdef _WAYNE_MPI
FILE
*f = myfopen(rellBootstrapFileNamePID, "wb");
#else
FILE
*f = myfopen(rellBootstrapFileName, "wb");
#endif
for(i = 0; i < NUM_RELL_BOOTSTRAPS; i++)
{
restoreTreeList(tr->rellTrees, tr, i);
Tree2String(tr->tree_string, tr, tr->start->back, FALSE, TRUE, FALSE, FALSE, TRUE, adef, SUMMARIZE_LH, FALSE, FALSE, FALSE, FALSE);
fprintf(f, "%s", tr->tree_string);
}
freeTreeList(tr->rellTrees);
rax_free(tr->rellTrees);
rax_free(tr->resample);
fclose(f);
#ifdef _WAYNE_MPI
MPI_Barrier(MPI_COMM_WORLD);
concatenateBSFiles(processes, rellBootstrapFileName);
removeBSFiles(processes, rellBootstrapFileName);
MPI_Barrier(MPI_COMM_WORLD);
if(processID == 0)
printBothOpen("\nRELL bootstraps written to file %s\n", rellBootstrapFileName);
#else
printBothOpen("\nRELL bootstraps written to file %s\n", rellBootstrapFileName);
#endif
}
#ifdef _WAYNE_MPI
if(processID == bestProcess)
{
#endif
overallTime = gettime() - masterTime;
printBothOpen("Program execution info written to %s\n", infoFileName);
if(!tr->catOnly)
{
printBothOpen("Best-scoring ML tree written to: %s\n\n", bestTreeFileName);
if(adef->perGeneBranchLengths && tr->NumberOfModels > 1)
printBothOpen("Per-Partition branch lengths of best-scoring ML tree written to %s.PARTITION.0 to %s.PARTITION.%d\n\n", bestTreeFileName, bestTreeFileName,
tr->NumberOfModels - 1);
}
printBothOpen("Overall execution time: %f secs or %f hours or %f days\n\n", overallTime, overallTime/3600.0, overallTime/86400.0);
#ifdef _WAYNE_MPI
}
#endif
if(!tr->catOnly)
{
freeTL(rl);
rax_free(rl);
}
#ifdef _WAYNE_MPI
MPI_Finalize();
#endif
exit(0);
}
void fastSearch(tree *tr, analdef *adef, rawdata *rdta, cruncheddata *cdta)
{
double
likelihood,
startLikelihood,
*lhVectors[3];
char
bestTreeFileName[1024];
FILE
*f;
int
model;
lhVectors[0] = (double *)NULL;
lhVectors[1] = (double *)NULL;
lhVectors[2] = (double *)NULL;
/* initialize model parameters with standard starting values */
initModel(tr, rdta, cdta, adef);
printBothOpen("Time after init : %f\n", gettime() - masterTime);
/*
compute starting tree, either by reading in a tree specified via -t
or by building one
*/
getStartingTree(tr, adef);
printBothOpen("Time after init and starting tree: %f\n", gettime() - masterTime);
/*
rough model parameter optimization, the log likelihood epsilon should
actually be determined based on the initial tree score and not be hard-coded
*/
if(adef->useBinaryModelFile)
{
readBinaryModel(tr);
evaluateGenericInitrav(tr, tr->start);
treeEvaluate(tr, 2);
}
else
modOpt(tr, adef, FALSE, 10.0);
printBothOpen("Time after init, starting tree, mod opt: %f\n", gettime() - masterTime);
/* print out the number of rate categories used for the CAT model, one should
use less then the default, e.g., -c 16 works quite well */
for(model = 0; model < tr->NumberOfModels; model++)
printBothOpen("Partion %d number of Cats: %d\n", model, tr->partitionData[model].numberOfCategories);
/*
means that we are going to do thorough insertions
with real newton-raphson based br-len opt at the three branches
adjactent to every insertion point
*/
Thorough = 1;
/*
loop over SPR cycles until the likelihood difference
before and after the SPR cycle is <= 0.5 log likelihood units.
Rather than being hard-coded this should also be determined based on the
actual likelihood of the tree
*/
do
{
startLikelihood = tr->likelihood;
/* conduct a cycle of linear SPRs */
likelihood = linearSPRs(tr, 20, adef->veryFast);
evaluateGeneric(tr, tr->start);
/* the NNIs also optimize br-lens of resulting topology a bit */
encapsulateNNIs(tr, lhVectors, FALSE);
printBothOpen("LH after SPRs %f, after NNI %f\n", likelihood, tr->likelihood);
}
while(ABS(tr->likelihood - startLikelihood) > 0.5);
/* print out the resulting tree to the RAxML_bestTree. file.
note that boosttrapping or doing multiple inferences won't work.
This thing computes a single tree and that's it */
strcpy(bestTreeFileName, workdir);
strcat(bestTreeFileName, "RAxML_fastTree.");
strcat(bestTreeFileName, run_id);
Tree2String(tr->tree_string, tr, tr->start->back, FALSE, TRUE, FALSE, FALSE, FALSE, adef, SUMMARIZE_LH, FALSE, FALSE);
f = myfopen(bestTreeFileName, "wb");
fprintf(f, "%s", tr->tree_string);
fclose(f);
printBothOpen("RAxML fast tree written to file: %s\n", bestTreeFileName);
writeBinaryModel(tr);
printBothOpen("Total execution time: %f\n", gettime() - masterTime);
printBothOpen("Good bye ... \n");
}
