KmerDistribution BWTAlgorithms::sampleKmerCounts(size_t kmerSize, size_t sampleSize, const BWT* pBWT)
{
// Learn k-mer occurrence distribution for this value of k
KmerDistribution distribution;
for(size_t i = 0; i < sampleSize; ++i)
{
std::string s = sampleRandomString(pBWT, kmerSize);
int count = countSequenceOccurrences(s, pBWT);
distribution.add(count);
}
return distribution;
}
// Learn parameters of the kmer corrector
int learnKmerParameters(const BWT* pBWT)
{
std::cout << "Learning kmer parameters\n";
srand(time(0));
size_t n_samples = 10000;
//
KmerDistribution kmerDistribution;
int k = opt::kmerLength;
for(size_t i = 0; i < n_samples; ++i)
{
std::string s = BWTAlgorithms::sampleRandomString(pBWT);
int n = s.size();
int nk = n - k + 1;
for(int j = 0; j < nk; ++j)
{
std::string kmer = s.substr(j, k);
int count = BWTAlgorithms::countSequenceOccurrences(kmer, pBWT);
kmerDistribution.add(count);
}
}
//
kmerDistribution.print(75);
double ratio = 2.0f;
int chosenThreshold = kmerDistribution.findErrorBoundaryByRatio(ratio);
double cumulativeLEQ = kmerDistribution.getCumulativeProportionLEQ(chosenThreshold);
if(chosenThreshold == -1)
{
std::cerr << "[sga correct] Error k-mer threshold learning failed\n";
std::cerr << "[sga correct] This can indicate the k-mer you choose is too high or your data has very low coverage\n";
exit(EXIT_FAILURE);
}
std::cout << "Chosen kmer threshold: " << chosenThreshold << "\n";
std::cout << "Proportion of kmer density right of threshold: " << 1.0f - cumulativeLEQ << "\n";
if(cumulativeLEQ > 0.25f)
{
std::cerr << "[sga correct] Warning: Proportion of kmers greater than the chosen threshold is less than 0.75 (" << cumulativeLEQ << "\n";
std::cerr << "[sga correct] This can indicate your chosen kmer size is too large or your data is too low coverage to reliably correct\n";
std::cerr << "[sga correct] It is suggest to lower the kmer size and/or choose the threshold manually\n";
}
return chosenThreshold;
}
void generate_kmer_coverage(JSONWriter* pWriter, const BWTIndexSet& index_set)
{
pWriter->String("KmerDistribution");
pWriter->StartObject();
size_t n_samples = 10000;
size_t k = 51;
pWriter->String("k");
pWriter->Int(k);
KmerDistribution kmerDistribution;
for(size_t i = 0; i < n_samples; ++i)
{
std::string s = BWTAlgorithms::sampleRandomString(index_set.pBWT);
int n = s.size();
int nk = n - k + 1;
for(int j = 0; j < nk; ++j)
{
std::string kmer = s.substr(j, k);
int count = BWTAlgorithms::countSequenceOccurrences(kmer, index_set.pBWT);
kmerDistribution.add(count);
}
}
pWriter->String("distribution");
pWriter->StartArray();
int max = kmerDistribution.getCutoffForProportion(0.95f);
std::vector<int> count_vector = kmerDistribution.toCountVector(max);
for(size_t i = 1; i < count_vector.size(); ++i)
{
pWriter->StartObject();
pWriter->String("kmer-depth");
pWriter->Int(i);
pWriter->String("count");
pWriter->Int(count_vector[i]);
pWriter->EndObject();
}
pWriter->EndArray();
pWriter->EndObject();
}