/// <summary>
/// Verifies the alignment has no leading or trailing gaps and throws an exception otherwise.
/// </summary>
/// <param name="refseq">Refseq.</param>
/// <param name="query">Query.</param>
internal static void VerifyNoGapsOnEnds(byte[] refseq, BPandQV[] query) {
var gap = DnaAlphabet.Instance.Gap;
if (refseq [0] == gap ||
refseq [refseq.Length - 1] == gap ||
query [0].BP == gap ||
query [query.Length - 1].BP == gap) {
var refseqs = new string(refseq.Select(x=>(char)x).ToArray());
var qseqs = new string(query.Select(x=>(char)x.BP).ToArray());
throw new FormatException ("Alignment query and/or reference started with a gap character. " +
"Alignments must be hard-clipped to remove starting and trailing variants " +
"before variants can be called. Alignment was:\n" + refseqs + "\n" + qseqs);
}
}
/// <summary>
/// Reverse complements the BP and QV values accounting for homopolymers.
///
/// This is not a simple operation because in addition to reversing the QV scores, one must account for the fact that
/// the quality value for homopolymer indel errors (that is a deletion or insertion) is typically only placed at the
/// first base in a homopolymer (though this is not standardized). To account for this, when reverse complementing, we switch the
/// QV value of the first and last base in homopolymers if the first base is lower quality than the last base.
/// </summary>
/// <returns>The reverse complemented sequence.</returns>
/// <param name="toFlip">The array with the QV values to flip.</param>
/// <param name="flipHpQvValues">If set to <c>true</c> flip hp qv values.</param>
internal static BPandQV[] GetReverseComplementedSequence(BPandQV[] toFlip, bool flipHpQvValues = false)
{
BPandQV[] newData = new BPandQV[toFlip.Length];
for (long index = 0; index < toFlip.Length; index++)
{
byte complementedSymbol;
byte symbol = toFlip[toFlip.Length - index - 1].BP;
if (!DnaAlphabet.Instance.TryGetComplementSymbol(symbol, out complementedSymbol))
{
throw new NotSupportedException("Bad character in BPandQV array: " + symbol.ToString());
}
var bpandq = new BPandQV(complementedSymbol, toFlip[toFlip.Length - index - 1].QV);
newData [index] = bpandq;
}
if (flipHpQvValues)
{
ReverseQVValuesForHomopolymers(newData);
}
return(newData);
}
/// <summary>
/// Gets the bases for a length of the sequence as a string.
/// </summary>
/// <returns>The bases.</returns>
/// <param name="array">Array.</param>
/// <param name="position">Position.</param>
/// <param name="length">Length.</param>
private static string getBases(BPandQV[] array, int position, int length)
{
char[] chars = new char[length];
for(int i=0; i<length; i++)
{
chars[i] = (char)array[i+position].BP;
}
return new string(chars);
}
/// <summary>
/// Calls the variants.
///
/// Should only be used internally as assumptions are made that the alignments are left-aligned and fulfill certain criteria.
/// </summary>
/// <returns>The variants.</returns>
/// <param name="refSeq">Reference seq.</param>
/// <param name="querySeq">Query seq.</param>
/// <param name="originallyReverseComplemented">If set to <c>true</c> the query sequence was originally reverse complemented. (this affects QV value scoring)</param>
internal static List<Variant> CallVariants(byte[] refSeq, BPandQV[] querySeq, bool originallyReverseComplemented)
{
if (originallyReverseComplemented) {
AlignmentUtils.ReverseQVValuesForHomopolymers (querySeq);
}
List<Variant> variants = new List<Variant>();
// Now call variants.
var gap = DnaAlphabet.Instance.Gap;
int i = 0;
int refPos = 0;
while( i < refSeq.Length)
{
if (refSeq[i] == gap)
{
int len = AlignmentUtils.GetGapLength(i, refSeq);
var nextBasePos = (i + len);
// Should alway be true as we don't end in gaps
Debug.Assert (nextBasePos < refSeq.Length);
var hplenAndChar = determineHomoPolymerLength (nextBasePos, refSeq);
var bases = getBases(querySeq, i, len);
var newVariant = new IndelVariant(refPos - 1, len, bases, IndelType.Insertion,
hplenAndChar.Item2, hplenAndChar.Item1,
(i == 0 || (i + len + hplenAndChar.Item1) >= refSeq.Length));
newVariant.QV = querySeq[i].QV;
variants.Add(newVariant);
i += len;
}
else if (querySeq[i].BP == gap)
{
int len = AlignmentUtils.GetGapLength(i, querySeq);
var bases = getBases(refSeq, i, len);
var hplenAndChar = determineHomoPolymerLength (i, refSeq);
var newVariant = new IndelVariant(refPos - 1, len, bases,
IndelType.Deletion, hplenAndChar.Item2,
hplenAndChar.Item1, (i == 0 || (i + len + hplenAndChar.Item1) >= refSeq.Length));
/* An insertion mutation occurs BEFORE pos, so normally we get the next base
* or the last one if it's a reverse complemented alignment. However, this is not true if
* it is a homopolymer because what would have been the previous position is the next position
* after left aligning and reversing the position of the QV value.
*
* Consider the following
* --*- -*--
* A-TA --> TA-T
* AGTA TACT
*
* However,
* --*-- --*--
* A-TTA ----> T-AAT
* ATTTA TAAAT
*
*/
if ((i + len ) < querySeq.Length) {
var qc_pos = originallyReverseComplemented ? i - 1 : i + len;
if (newVariant.InHomopolymer) {
qc_pos = i + len;
}
newVariant.QV = querySeq[qc_pos].QV;
}
variants.Add(newVariant);
i += len;
refPos += len;
}
else
{
if (querySeq[i].BP != refSeq[i])
{
var newVariant = new SNPVariant(refPos, (char) querySeq[i].BP, (char)refSeq[i], (i ==0 || i == (refSeq.Length -1)));
newVariant.QV = querySeq [i].QV;
variants.Add(newVariant);
}
i++; refPos++;
}
}
return variants;
}
/// <summary>
/// Given two byte arrays representing a pairwise alignment, shift them so
/// that all deletions start as early as possible. For example:
/// TTTTAAAATTTT -> Converts to -> TTTTAAAATTTT
/// TTTTAA--TTTT TTTT--AATTTT
///
/// This modifies the array in place.
/// </summary>
/// <param name="refseq">Reference Sequency</param>
/// <param name="query">Query Sequence</param>
/// <returns></returns>
public static void LeftAlignIndels(byte[] refseq, BPandQV[] query)
{
// Validation
if (refseq.Length != query.Length)
{
throw new ArgumentException("Alignment passed to LeftAlignIndels had unequal length sequences");
}
ValidateNoOverlappingGaps(refseq, query);
byte gap = DnaAlphabet.Instance.Gap;
// Keep left aligning until we can't anymore, this is a
// do while loop because some downstream left alignments open up
// further ones upstream, even though this is rare.
int change_count = 0;
int loopsThrough = 0;
do
{
loopsThrough++;
change_count = 0;
for (int i = 1; i < refseq.Length; i++)
{
if (refseq[i] == gap)
{
int len = GetGapLength(i, refseq);
int left_side = i - 1;
int right_side = i - 1 + len;
while (left_side >= 0 && refseq[left_side] != gap && (refseq[left_side] == query[right_side].BP))
{
// Move the gap left.
if (right_side < refseq.Length)
{
refseq[right_side] = refseq[left_side];
}
refseq[left_side] = gap;
left_side--;
right_side--;
change_count++;
}
if (loopsThrough > MAX_LOOPS)
{
throw new Exception(MAX_LOOPS_ERROR);
}
}
else if (query[i].BP == gap)
{
int len = GetGapLength(i, query);
int left_side = i - 1;
int right_side = i - 1 + len;
while (left_side >= 0 && query[left_side].BP != gap && (query[left_side].BP == refseq[right_side]))
{
// Move the gap left.
if (right_side < query.Length)
{
query[right_side] = query[left_side];
}
query[left_side] = new BPandQV(gap, 0);
left_side--;
right_side--;
change_count++;
}
if (loopsThrough > MAX_LOOPS)
{
throw new Exception(MAX_LOOPS_ERROR);
}
}
}
} while (change_count > 0);
}
/// <summary>
/// Given two byte arrays representing a pairwise alignment, shift them so
/// that all deletions start as early as possible. For example:
/// TTTTAAAATTTT -> Converts to -> TTTTAAAATTTT
/// TTTTAA--TTTT TTTT--AATTTT
///
/// This modifies the array in place.
/// </summary>
/// <param name="refseq">Reference Sequency</param>
/// <param name="query">Query Sequence</param>
/// <returns></returns>
public static void LeftAlignIndels(byte[] refseq, BPandQV[] query)
{
// Validation
if (refseq.Length != query.Length) {
throw new ArgumentException("Alignment passed to LeftAlignIndels had unequal length sequences");
}
ValidateNoOverlappingGaps (refseq, query);
byte gap = DnaAlphabet.Instance.Gap;
// Keep left aligning until we can't anymore, this is a
// do while loop because some downstream left alignments open up
// further ones upstream, even though this is rare.
int change_count = 0;
int loopsThrough = 0;
do
{
loopsThrough++;
change_count = 0;
for (int i = 1; i < refseq.Length; i++)
{
if (refseq[i] == gap)
{
int len = GetGapLength(i, refseq);
int left_side = i - 1;
int right_side = i - 1 + len;
while (left_side >= 0 && refseq[left_side] != gap && (refseq[left_side] == query[right_side].BP))
{
// Move the gap left.
if (right_side < refseq.Length) {
refseq[right_side] = refseq[left_side];
}
refseq[left_side] = gap;
left_side--;
right_side--;
change_count++;
}
if (loopsThrough > MAX_LOOPS) {
throw new Exception(MAX_LOOPS_ERROR);
}
}
else if (query[i].BP == gap)
{
int len = GetGapLength(i, query);
int left_side = i - 1;
int right_side = i - 1 + len;
while (left_side >= 0 && query[left_side].BP != gap && (query[left_side].BP == refseq[right_side]))
{
// Move the gap left.
if (right_side < query.Length) {
query[right_side] = query[left_side];
}
query[left_side] = new BPandQV(gap, 0);
left_side--;
right_side--;
change_count++;
}
if (loopsThrough > MAX_LOOPS) {
throw new Exception(MAX_LOOPS_ERROR);
}
}
}
} while (change_count > 0);
}
/// <summary>
/// Reverses the QV values for homopolymers.
/// This is not a simple operation because in addition to reversing the QV scores, one must account for the fact that
/// the quality value for homopolymer indel errors (that is a deletion or insertion) is typically only placed at the
/// first base in a homopolymer (though this is not standardized). To account for this, when reverse complementing, we switch the
/// QV value of the first and last base in homopolymers if the first base is lower quality than the last base.
/// </summary>
/// <returns>The QV values for homopolymers.</returns>
/// <param name="toFlip">To flip.</param>
internal static void ReverseQVValuesForHomopolymers(BPandQV[] toFlip) {
// Basic idea is to assume it is A, C, G, T alphabet and flip HP values
// Also assumes all low QV is due to HP deletion/insertion error.
if (toFlip.Length > 1) {
byte lastbp = toFlip [0].BP;
int firstPos = 0;
int curLength = 1;
for (int i = 1; i < toFlip.Length; i++) {
byte newbp = toFlip [i].BP;
if (newbp != lastbp) {
if (curLength > 1) {
var right = toFlip [i - 1];
var left = toFlip[firstPos];
Debug.Assert (right.BP == left.BP);
if (right.QV < left.QV) {
toFlip [i - 1] = left;
toFlip [firstPos] = right;
}
}
firstPos = i;
lastbp = newbp;
curLength = 1;
} else if (newbp == lastbp) {
curLength++;
}
}
// Finally flip the end
if (curLength > 1) {
var tmp = toFlip [toFlip.Length - 1];
toFlip [toFlip.Length - 1] = toFlip [firstPos];
toFlip [firstPos] = tmp;
}
}
}
/// <summary>
/// Reverse complements the BP and QV values accounting for homopolymers.
///
/// This is not a simple operation because in addition to reversing the QV scores, one must account for the fact that
/// the quality value for homopolymer indel errors (that is a deletion or insertion) is typically only placed at the
/// first base in a homopolymer (though this is not standardized). To account for this, when reverse complementing, we switch the
/// QV value of the first and last base in homopolymers if the first base is lower quality than the last base.
/// </summary>
/// <returns>The reverse complemented sequence.</returns>
/// <param name="toFlip">The array with the QV values to flip.</param>
/// <param name="flipHpQvValues">If set to <c>true</c> flip hp qv values.</param>
internal static BPandQV[] GetReverseComplementedSequence(BPandQV[] toFlip, bool flipHpQvValues = false)
{
BPandQV[] newData = new BPandQV[toFlip.Length];
for (long index = 0; index < toFlip.Length; index++)
{
byte complementedSymbol;
byte symbol = toFlip[toFlip.Length - index - 1].BP;
if (!DnaAlphabet.Instance.TryGetComplementSymbol(symbol, out complementedSymbol))
{
throw new NotSupportedException("Bad character in BPandQV array: " + symbol.ToString());
}
var bpandq = new BPandQV(complementedSymbol, toFlip[toFlip.Length - index -1].QV);
newData [index] = bpandq;
}
if (flipHpQvValues) {
ReverseQVValuesForHomopolymers (newData);
}
return newData;
}
/// <summary>
/// Simple check that the alignment does not have a gap on top of a
/// gap, which violates several assumptions.
/// </summary>
/// <param name="seq1"></param>
/// <param name="seq2"></param>
internal static void ValidateNoOverlappingGaps(byte[] seq1, BPandQV[] seq2)
{
var gap = DnaAlphabet.Instance.Gap;
for(int i=0;i<seq1.Length;i++)
{
if (seq1[i] == gap && seq2[i].BP == gap)
throw new Exception("You have an alignment with overlapping gaps. Input problem!");
}
}
/// <summary>
/// Given the start position of a gap, returns how long it is.
/// For example:
///
/// AAAA---TTTT returns 3.
/// </summary>
/// <param name="pos">0 indexed</param>
/// <param name="array"></param>
/// <returns></returns>
public static int GetGapLength(int pos, BPandQV[] array)
{
var gap = DnaAlphabet.Instance.Gap;
int len = 1;
while (++pos < array.Length)
{
if (array[pos].BP == gap)
{
len += 1;
}
else
{
break;
}
}
return len;
}