public static List <SequenceAlignment> GetFilteredSequenceAlignments(IChain n, bool nKeepRepeats, IChain c, bool cKeepRepeats, SS allowedTypes, int minAlignmentLength = DefaultMinAlignmentLength, float maxRmsd = DefaultRmsdThreshold)
{
List <SequenceAlignment> alignments = GetAlignmentsPreservingFullSsBlocks(n, c, allowedTypes, minAlignmentLength, maxRmsd);
int maxThirdHelixShortening = DefaultMaxThirdHelixTrimming;
//int maxOvershoot = 8;
// Find all alignments and remove those that would:
// -- remove more than one repeat of a repeat protein
if (nKeepRepeats)
{
int nRepeatLength;
if (Sequence.TryGetInternalRepeatLength(n, out nRepeatLength) && nRepeatLength > 15)
{
alignments.RemoveAll(a => a.Range1.End < n.Count - 1.25 * nRepeatLength);
}
}
if (cKeepRepeats)
{
int cRepeatLength;
if (Sequence.TryGetInternalRepeatLength(c, out cRepeatLength) && cRepeatLength > 15)
{
alignments.RemoveAll(a => a.Range1.Start < 1.25 * cRepeatLength);
}
}
// -- leave two or fewer secondary structure elements in either chain. An exception is made if there are only two secondary structure
// elements total to begin with, which is common for helical bundles
List <SSBlock> nBlocks = SecondaryStructure.GetPhiPsiSSBlocksOfType(n, SS.Helix | SS.Extended, minAlignmentLength);
List <SSBlock> cBlocks = SecondaryStructure.GetPhiPsiSSBlocksOfType(c, SS.Helix | SS.Extended, minAlignmentLength);
if (nBlocks.Count > 2)
{
alignments.RemoveAll(a => a.Range1.Start < nBlocks[1].End);
}
if (cBlocks.Count > 2)
{
alignments.RemoveAll(a => a.Range2.End > cBlocks[cBlocks.Count - 2].Start);
}
// -- shorten the third helix of a 3-helix bundle too much, because that would also probably destabilize the oligomer
if (nBlocks.Count > 2)
{
int removeCount = alignments.RemoveAll(a => a.Range1.End < nBlocks[2].End - maxThirdHelixShortening);
}
if (cBlocks.Count > 2)
{
int removeCount = alignments.RemoveAll(a => a.Range2.Start > cBlocks[cBlocks.Count - 3].Start + maxThirdHelixShortening);
}
return(alignments);
}
/// <summary>
/// Find splicing alignment ranges for two peptides that would retain a full secondary structure block for at least one of the two peptides. This prevents
/// short splice overlaps that remove some of the secondary structure on both sides of the alignment. Basically, the alignment must result in either:
/// 1) one secondary structure block being fully included in the other
/// 2) the N-terminal peptide being spliced has the end of its block fully in the resultant alignment, thus preserving the entire N-side block
/// 3) the C-terminal peptide being spliced has the start of its block fully in the resultant alignment, thus preserving the entire C-side block
/// </summary>
/// <param name="n"></param>
/// <param name="c"></param>
/// <param name="allowedTypes"></param>
/// <param name="minAlignmentLength"></param>
/// <returns></returns>
public static List <SequenceAlignment> GetAlignmentsPreservingFullSsBlocks(IChain n, IChain c, SS allowedTypes, int minAlignmentLength = DefaultMinAlignmentLength, float maxRmsd = DefaultRmsdThreshold)
{
// TODO!!!: Debug why calling this function with peptides A,B vs B,A returns different numbers of alignments.
List <SequenceAlignment> alignments = new List <SequenceAlignment>();
List <SSBlock> nBlocks = SecondaryStructure.GetPhiPsiSSBlocksOfType(n, allowedTypes, minAlignmentLength).Where(block => block.Length >= minAlignmentLength).ToList();
List <SSBlock> cBlocks = SecondaryStructure.GetPhiPsiSSBlocksOfType(c, allowedTypes, minAlignmentLength).Where(block => block.Length >= minAlignmentLength).ToList();
foreach (SSBlock nBlock in nBlocks)
{
foreach (SSBlock cBlock in cBlocks)
{
for (int nAlignmentOfBlockC = nBlock.End - minAlignmentLength + 1; nAlignmentOfBlockC + cBlock.Length > nBlock.Start + minAlignmentLength; nAlignmentOfBlockC--)
{
// Figure out the start and ends of the overlap region
int nStart = Math.Max(nBlock.Start, nAlignmentOfBlockC);
int nEnd = Math.Min(nBlock.End, nAlignmentOfBlockC + cBlock.Length - 1);
int alignmentLength = nEnd - nStart + 1;
int cStart = Math.Max(cBlock.Start, cBlock.Start + (nStart - nAlignmentOfBlockC));
int cEnd = cStart + alignmentLength - 1;
Debug.Assert(nBlock.Start <= nStart && nEnd <= nBlock.End);
Debug.Assert(cBlock.Start <= cStart && cEnd <= cBlock.End);
Debug.Assert(nEnd - nStart == cEnd - cStart);
bool cFullyIncluded = (cBlock.Start == cStart && cBlock.End == cEnd);
bool nFullyIncluded = (nBlock.Start == nStart && nBlock.End == nEnd);
if (!nFullyIncluded && !cFullyIncluded && cStart != cBlock.Start && nEnd != nBlock.End)
{
continue;
}
if (Rmsd.GetRmsdNCAC(n, nStart, nEnd, c, cStart, cEnd) > maxRmsd)
{
continue;
}
alignments.Add(new SequenceAlignment(nStart, nEnd, cStart, cEnd));
}
}
}
return(alignments);
}
/// <summary>
/// Find splicing alignment ranges for two peptides that would retain a full secondary structure block for at least one of the two peptides. This prevents
/// short splice overlaps that remove some of the secondary structure on both sides of the alignment. Basically, the alignment must result in either:
/// 1) one secondary structure block being fully included in the other
/// 2) the N-terminal peptide being spliced has the end of its block fully in the resultant alignment, thus preserving the entire N-side block
/// 3) the C-terminal peptide being spliced has the start of its block fully in the resultant alignment, thus preserving the entire C-side block
/// </summary>
/// <param name="chain1"></param>
/// <param name="chain2"></param>
/// <param name="allowedTypes"></param>
/// <param name="minAlignmentLength"></param>
/// <returns></returns>
public static List <TransformSequenceAlignment> GetTransformAlignments(IChain chain1, IChain chain2, SS allowedTypes, int minAlignmentLength = DefaultMinAlignmentLength, float maxRmsd = DefaultRmsdThreshold)
{
List <TransformSequenceAlignment> alignments = new List <TransformSequenceAlignment>();
List <SSBlock> nBlocks = SecondaryStructure.GetPhiPsiSSBlocksOfType(chain1, allowedTypes, minAlignmentLength).Where(block => block.Length >= minAlignmentLength).ToList();
List <SSBlock> cBlocks = SecondaryStructure.GetPhiPsiSSBlocksOfType(chain2, allowedTypes, minAlignmentLength).Where(block => block.Length >= minAlignmentLength).ToList();
// Previously, calling this function with peptides A,B vs B,A returns different numbers of alignments. The original intent was for the first chain passed in to be at the
// N-terminus and the second at the C-terminus (post-fusion), so these post-fusion lengths were being calculated here and culled differently based on order.
foreach (SSBlock nBlock in nBlocks)
{
foreach (SSBlock cBlock in cBlocks)
{
for (int nAlignmentOfBlockC = nBlock.End - minAlignmentLength + 1; nAlignmentOfBlockC + cBlock.Length >= nBlock.Start + minAlignmentLength; nAlignmentOfBlockC--)
{
// Figure out the start and ends of the overlap region
int nStart = Math.Max(nBlock.Start, nAlignmentOfBlockC);
int nEnd = Math.Min(nBlock.End, nAlignmentOfBlockC + cBlock.Length - 1);
int alignmentLength = nEnd - nStart + 1;
int cStart = Math.Max(cBlock.Start, cBlock.Start + (nStart - nAlignmentOfBlockC));
int cEnd = cStart + alignmentLength - 1;
//Debug.WriteLine("Comparing {0}-{1} vs {2}-{3}", nStart, nEnd, cStart, cEnd);
Debug.Assert(nBlock.Start <= nStart && nEnd <= nBlock.End);
Debug.Assert(cBlock.Start <= cStart && cEnd <= cBlock.End);
Debug.Assert(nEnd - nStart == cEnd - cStart);
//bool cFullyIncluded = (cBlock.Start == cStart && cBlock.End == cEnd);
//bool nFullyIncluded = (nBlock.Start == nStart && nBlock.End == nEnd);
//if (!nFullyIncluded && !cFullyIncluded && cStart != cBlock.Start && nEnd != nBlock.End)
// continue;
Trace.Assert(nEnd - nStart + 1 >= minAlignmentLength);
float rmsd = float.NaN;
Matrix matrix = Rmsd.GetRmsdAndTransform(chain1, nStart, nEnd, chain2, cStart, cEnd, out rmsd);
bool fail = rmsd > maxRmsd;
#if DEBUG && false
float rmsd2 = float.NaN;
Matrix matrix2 = Rmsd.GetRmsdAndTransform(chain2, cStart, cEnd, chain1, nStart, nEnd, out rmsd2);
bool fail2 = rmsd2 > maxRmsd;
Trace.Assert(fail == fail2);
#endif
if (rmsd > maxRmsd)
{
continue;
}
TransformSequenceAlignment alignment = new TransformSequenceAlignment(nStart, nEnd, cStart, cEnd);
alignment.Centroid1 = Geometry.GetCenterNCAC(chain1[nStart, nEnd]);
alignment.Centroid2 = Geometry.GetCenterNCAC(chain2[cStart, cEnd]);
alignment.Align1 = matrix;
alignment.Align2 = Matrix.Invert(matrix);
alignments.Add(alignment);
}
}
}
return(alignments);
}
public static IChain GetChain(IChain[] peptides, SequenceAlignment[] alignments, Selection immutableAas)
{
IChain fusion = new Chain();
// Do all pairwise analysis
for (int i = 0; i < peptides.Length - 1; i++)
{
// Determine the ranges outside of the splice
int start1 = i == 0 ? 0 : alignments[i - 1].Range2.End + 1;
int end1 = alignments[i].Range1.Start - 1;
int start2 = alignments[i].Range2.End + 1;
int end2 = i < alignments.Length - 1 ? alignments[i + 1].Range1.Start - 1 : peptides[i + 1].Count - 1;
// Add the non-overlapping region of the first peptide
if (start1 <= end1)
{
foreach (Aa aa in peptides[i][start1, end1])
{
Aa copy = new Aa(aa, i == 0 && start1 == 0, false);
copy.NodeTransform = aa.TotalTransform;
fusion.Add(copy);
}
}
// Add the alignment region, selecting either from the first or second peptide so as to minimize clashes with the sidechains that
// are for sure being kept on either side
SequenceAlignment alignment = alignments[i];
Debug.Assert(alignment.Range1.Length == alignment.Range2.Length);
for (int alignmentOffset = 0; alignmentOffset < alignment.Range1.Length; alignmentOffset++)
{
int index1 = alignment.Range1.Start + alignmentOffset;
int index2 = alignment.Range2.Start + alignmentOffset;
IAa option1 = peptides[i][index1];
IAa option2 = peptides[i + 1][index2];
bool nTerminus = i == 0 && index1 == 0;
bool cTerminus = (i == peptides.Length - 2) && (index2 == peptides[i + 1].Count - 1);
if (immutableAas.Aas.Contains(option1))
{
Aa copy = new Aa(option1, nTerminus, cTerminus);
copy.NodeTransform = option1.TotalTransform;
fusion.Add(copy);
}
else if (immutableAas.Aas.Contains(option2))
{
Aa copy = new Aa(option2, nTerminus, cTerminus);
copy.NodeTransform = option2.TotalTransform;
fusion.Add(copy);
}
else
{
if (option2.Letter == 'P' && index1 >= 4)
{
SS[] ss1 = SecondaryStructure.GetPhiPsiSS(peptides[i], 5);
bool allHelical = (ss1[index1] | ss1[index1 - 1] | ss1[index1 - 2] | ss1[index1 - 3] | ss1[index1 - 4]) == SS.Helix;
if (allHelical)
{
Aa copy = new Aa(option1, nTerminus, cTerminus);
copy.NodeTransform = option1.TotalTransform;
fusion.Add(copy);
continue;
}
}
// Otherwise, select the residue with fewer clashes
int clashCount1 = end2 >= start2? peptides[i + 1][start2, end2].Select(other => Clash.AnyContact(other, option1, Clash.ContactType.SidechainSidechainClash) ? 1 : 0).Aggregate(0, (a, b) => a + b) : 0;
int clashCount2 = end1 >= start1? peptides[i][start1, end1].Select(other => Clash.AnyContact(other, option2, Clash.ContactType.SidechainSidechainClash) ? 1 : 0).Aggregate(0, (a, b) => a + b) : 0;
if (clashCount1 <= clashCount2)
{
Aa copy = new Aa(option1, nTerminus, cTerminus);
copy.NodeTransform = option1.TotalTransform;
fusion.Add(copy);
continue;
}
if (clashCount2 < clashCount1)
{
Aa copy = new Aa(option2, nTerminus, cTerminus);
copy.NodeTransform = option2.TotalTransform;
fusion.Add(copy);
continue;
}
}
}
// Add the non-overlapping region of the last peptide
if (i == peptides.Length - 2 && start2 <= end2)
{
foreach (Aa aa in peptides[i + 1][start2, end2])
{
Aa copy = new Aa(aa, false, aa.IsCTerminus);
copy.NodeTransform = aa.TotalTransform;
fusion.Add(copy);
}
}
}
return(fusion);
}