*/
public MapList()
{
- fromShifts = new ArrayList<int[]>();
- toShifts = new ArrayList<int[]>();
+ fromShifts = new ArrayList<>();
+ toShifts = new ArrayList<>();
}
/**
{
return false;
}
- return Arrays
- .deepEquals(fromShifts.toArray(), obj.fromShifts.toArray())
- && Arrays
- .deepEquals(toShifts.toArray(), obj.toShifts.toArray());
+ return Arrays.deepEquals(fromShifts.toArray(), obj.fromShifts.toArray())
+ && Arrays.deepEquals(toShifts.toArray(),
+ obj.toShifts.toArray());
}
/**
{
int hashCode = 31 * fromRatio;
hashCode = 31 * hashCode + toRatio;
- hashCode = 31 * hashCode + fromShifts.toArray().hashCode();
- hashCode = 31 * hashCode + toShifts.toArray().hashCode();
+ for (int[] shift : fromShifts)
+ {
+ hashCode = 31 * hashCode + shift[0];
+ hashCode = 31 * hashCode + shift[1];
+ }
+ for (int[] shift : toShifts)
+ {
+ hashCode = 31 * hashCode + shift[0];
+ hashCode = 31 * hashCode + shift[1];
+ }
+
return hashCode;
}
/**
* Constructor given from and to ranges as [start1, end1, start2, end2,...].
- * If any end is equal to the next start, the ranges will be merged. There is
- * no validation check that the ranges do not overlap each other.
+ * There is no validation check that the ranges do not overlap each other.
*
* @param from
* contiguous regions as [start1, end1, start2, end2, ...]
this.toRatio = toRatio;
fromLowest = Integer.MAX_VALUE;
fromHighest = Integer.MIN_VALUE;
- int added = 0;
for (int i = 0; i < from.length; i += 2)
{
*/
fromLowest = Math.min(fromLowest, Math.min(from[i], from[i + 1]));
fromHighest = Math.max(fromHighest, Math.max(from[i], from[i + 1]));
- if (added > 0 && from[i] == fromShifts.get(added - 1)[1])
- {
- /*
- * this range starts where the last ended - just extend it
- */
- fromShifts.get(added - 1)[1] = from[i + 1];
- }
- else
- {
- fromShifts.add(new int[] { from[i], from[i + 1] });
- added++;
- }
+ fromShifts.add(new int[] { from[i], from[i + 1] });
}
toLowest = Integer.MAX_VALUE;
toHighest = Integer.MIN_VALUE;
- added = 0;
for (int i = 0; i < to.length; i += 2)
{
toLowest = Math.min(toLowest, Math.min(to[i], to[i + 1]));
toHighest = Math.max(toHighest, Math.max(to[i], to[i + 1]));
- if (added > 0 && to[i] == toShifts.get(added - 1)[1])
- {
- toShifts.get(added - 1)[1] = to[i + 1];
- }
- else
- {
- toShifts.add(new int[] { to[i], to[i + 1] });
- added++;
- }
+ toShifts.add(new int[] { to[i], to[i + 1] });
}
}
fromHighest = Integer.MIN_VALUE;
for (int[] range : fromRange)
{
+ if (range.length != 2)
+ {
+ // throw new IllegalArgumentException(range);
+ System.err.println("Invalid format for fromRange "
+ + Arrays.toString(range) + " may cause errors");
+ }
fromLowest = Math.min(fromLowest, Math.min(range[0], range[1]));
fromHighest = Math.max(fromHighest, Math.max(range[0], range[1]));
}
toHighest = Integer.MIN_VALUE;
for (int[] range : toRange)
{
+ if (range.length != 2)
+ {
+ // throw new IllegalArgumentException(range);
+ System.err.println("Invalid format for toRange "
+ + Arrays.toString(range) + " may cause errors");
+ }
toLowest = Math.min(toLowest, Math.min(range[0], range[1]));
toHighest = Math.max(toHighest, Math.max(range[0], range[1]));
}
/**
* Consolidates a list of ranges so that any contiguous ranges are merged.
* This assumes the ranges are already in start order (does not sort them).
+ * <p>
+ * The main use case for this method is when mapping cDNA sequence to its
+ * protein product, based on CDS feature ranges which derive from spliced
+ * exons, but are contiguous on the cDNA sequence. For example
+ *
+ * <pre>
+ * CDS 1-20 // from exon1
+ * CDS 21-35 // from exon2
+ * CDS 36-71 // from exon3
+ * 'coalesce' to range 1-71
+ * </pre>
*
* @param ranges
* @return the same list (if unchanged), else a new merged list, leaving the
}
boolean changed = false;
- List<int[]> merged = new ArrayList<int[]>();
+ List<int[]> merged = new ArrayList<>();
int[] lastRange = ranges.get(0);
int lastDirection = lastRange[1] >= lastRange[0] ? 1 : -1;
lastRange = new int[] { lastRange[0], lastRange[1] };
first = false;
continue;
}
- if (range[0] == lastRange[0] && range[1] == lastRange[1])
- {
- // drop duplicate range
- changed = true;
- continue;
- }
-
- /*
- * drop this range if it lies within the last range
- */
- if ((lastDirection == 1 && range[0] >= lastRange[0]
- && range[0] <= lastRange[1] && range[1] >= lastRange[0] && range[1] <= lastRange[1])
- || (lastDirection == -1 && range[0] <= lastRange[0]
- && range[0] >= lastRange[1]
- && range[1] <= lastRange[0] && range[1] >= lastRange[1]))
- {
- changed = true;
- continue;
- }
int direction = range[1] >= range[0] ? 1 : -1;
boolean sameDirection = range[1] == range[0]
|| direction == lastDirection;
boolean extending = range[0] == lastRange[1] + lastDirection;
- boolean overlapping = (lastDirection == 1 && range[0] >= lastRange[0] && range[0] <= lastRange[1])
- || (lastDirection == -1 && range[0] <= lastRange[0] && range[0] >= lastRange[1]);
- if (sameDirection && (overlapping || extending))
+ if (sameDirection && extending)
{
lastRange[1] = range[1];
changed = true;
List<int[]> shiftFrom, int toRatio)
{
// TODO: javadoc; tests
- int[] fromCount = countPos(shiftTo, pos);
+ int[] fromCount = countPositions(shiftTo, pos);
if (fromCount == null)
{
return null;
}
int fromRemainder = (fromCount[0] - 1) % fromRatio;
int toCount = 1 + (((fromCount[0] - 1) / fromRatio) * toRatio);
- int[] toPos = countToPos(shiftFrom, toCount);
+ int[] toPos = traverseToPosition(shiftFrom, toCount);
if (toPos == null)
{
- return null; // throw new Error("Bad Mapping!");
+ return null;
}
- // System.out.println(fromCount[0]+" "+fromCount[1]+" "+toCount);
return new int[] { toPos[0], fromRemainder, toPos[1] };
}
/**
- * count how many positions pos is along the series of intervals.
+ * Counts how many positions pos is along the series of intervals. Returns an
+ * array of two values:
+ * <ul>
+ * <li>the number of positions traversed (inclusive) to reach {@code pos}</li>
+ * <li>+1 if the last interval traversed is forward, -1 if in a negative
+ * direction</li>
+ * </ul>
+ * Returns null if {@code pos} does not lie in any of the given intervals.
*
- * @param shiftTo
+ * @param intervals
+ * a list of start-end intervals
* @param pos
- * @return number of positions or null if pos is not within intervals
+ * a position that may lie in one (or more) of the intervals
+ * @return
*/
- protected static int[] countPos(List<int[]> shiftTo, int pos)
+ protected static int[] countPositions(List<int[]> intervals, int pos)
{
- int count = 0, intv[], iv = 0, ivSize = shiftTo.size();
+ int count = 0;
+ int iv = 0;
+ int ivSize = intervals.size();
+
while (iv < ivSize)
{
- intv = shiftTo.get(iv++);
+ int[] intv = intervals.get(iv++);
if (intv[0] <= intv[1])
{
+ /*
+ * forwards interval
+ */
if (pos >= intv[0] && pos <= intv[1])
{
return new int[] { count + pos - intv[0] + 1, +1 };
}
else
{
+ /*
+ * reverse interval
+ */
if (pos >= intv[1] && pos <= intv[0])
{
return new int[] { count + intv[0] - pos + 1, -1 };
}
/**
- * count out pos positions into a series of intervals and return the position
+ * Reads through the given intervals until {@code count} positions have been
+ * traversed, and returns an array consisting of two values:
+ * <ul>
+ * <li>the value at the {@code count'th} position</li>
+ * <li>+1 if the last interval read is forwards, -1 if reverse direction</li>
+ * </ul>
+ * Returns null if the ranges include less than {@code count} positions, or if
+ * {@code count < 1}.
*
- * @param shiftFrom
- * @param pos
- * @return position pos in interval set
+ * @param intervals
+ * a list of [start, end] ranges
+ * @param count
+ * the number of positions to traverse
+ * @return
*/
- protected static int[] countToPos(List<int[]> shiftFrom, int pos)
+ protected static int[] traverseToPosition(List<int[]> intervals,
+ final int count)
{
- int count = 0, diff = 0, iv = 0, ivSize = shiftFrom.size();
- int[] intv = { 0, 0 };
+ int traversed = 0;
+ int ivSize = intervals.size();
+ int iv = 0;
+
+ if (count < 1)
+ {
+ return null;
+ }
+
while (iv < ivSize)
{
- intv = shiftFrom.get(iv++);
- diff = intv[1] - intv[0];
+ int[] intv = intervals.get(iv++);
+ int diff = intv[1] - intv[0];
if (diff >= 0)
{
- if (pos <= count + 1 + diff)
+ if (count <= traversed + 1 + diff)
{
- return new int[] { pos - count - 1 + intv[0], +1 };
+ return new int[] { intv[0] + (count - traversed - 1), +1 };
}
else
{
- count += 1 + diff;
+ traversed += 1 + diff;
}
}
else
{
- if (pos <= count + 1 - diff)
+ if (count <= traversed + 1 - diff)
{
- return new int[] { intv[0] - (pos - count - 1), -1 };
+ return new int[] { intv[0] - (count - traversed - 1), -1 };
}
else
{
- count += 1 - diff;
+ traversed += 1 - diff;
}
}
}
- return null;// (diff<0) ? (intv[1]-1) : (intv[0]+1);
+ return null;
}
/**
*/
public int[] locateInFrom(int start, int end)
{
+ return locateInFrom2(start, end);
+
// inefficient implementation
- int fromStart[] = shiftTo(start);
+ // int fromStart[] = shiftTo(start);
// needs to be inclusive of end of symbol position
- int fromEnd[] = shiftTo(end);
-
- return getIntervals(fromShifts, fromStart, fromEnd, fromRatio);
+ // int fromEnd[] = shiftTo(end);
+ // return getIntervals(fromShifts, fromStart, fromEnd, fromRatio);
}
/**
*/
public int[] locateInTo(int start, int end)
{
- int toStart[] = shiftFrom(start);
- int toEnd[] = shiftFrom(end);
- return getIntervals(toShifts, toStart, toEnd, toRatio);
+ return locateInTo2(start, end);
+
+ // int toStart[] = shiftFrom(start);
+ // int toEnd[] = shiftFrom(end);
+ // return getIntervals(toShifts, toStart, toEnd, toRatio);
}
/**
{
return null;
}
- List<int[]> ranges = new ArrayList<int[]>();
+ List<int[]> ranges = new ArrayList<>();
if (fs <= fe)
{
intv = fs;
// TODO not used - remove?
if (local)
{
- return ((getFromLowest() >= map.getFromLowest() && getFromHighest() <= map
- .getFromHighest()) || (getFromLowest() <= map.getFromLowest() && getFromHighest() >= map
- .getFromHighest()));
+ return ((getFromLowest() >= map.getFromLowest()
+ && getFromHighest() <= map.getFromHighest())
+ || (getFromLowest() <= map.getFromLowest()
+ && getFromHighest() >= map.getFromHighest()));
}
else
{
- return ((getToLowest() >= map.getToLowest() && getToHighest() <= map
- .getToHighest()) || (getToLowest() <= map.getToLowest() && getToHighest() >= map
- .getToHighest()));
+ return ((getToLowest() >= map.getToLowest()
+ && getToHighest() <= map.getToHighest())
+ || (getToLowest() <= map.getToLowest()
+ && getToHighest() >= map.getToHighest()));
}
}
*/
public boolean isFromForwardStrand()
{
+ return isForwardStrand(getFromRanges());
+ }
+
+ /**
+ * Returns true if mapping is to forward strand, false if to reverse strand.
+ * Result is just based on the first 'to' range that is not a single position.
+ * Default is true unless proven to be false. Behaviour is not well defined if
+ * the mapping has a mixture of forward and reverse ranges.
+ *
+ * @return
+ */
+ public boolean isToForwardStrand()
+ {
+ return isForwardStrand(getToRanges());
+ }
+
+ /**
+ * A helper method that returns true unless at least one range has start >
+ * end. Behaviour is undefined for a mixture of forward and reverse ranges.
+ *
+ * @param ranges
+ * @return
+ */
+ private boolean isForwardStrand(List<int[]> ranges)
+ {
boolean forwardStrand = true;
- for (int[] range : getFromRanges())
+ for (int[] range : ranges)
{
if (range[1] > range[0])
{
|| (fromRatio == 3 && toRatio == 1);
}
+ /**
+ * Returns a map which is the composite of this one and the input map. That
+ * is, the output map has the fromRanges of this map, and its toRanges are the
+ * toRanges of this map as transformed by the input map.
+ * <p>
+ * Returns null if the mappings cannot be traversed (not all toRanges of this
+ * map correspond to fromRanges of the input), or if this.toRatio does not
+ * match map.fromRatio.
+ *
+ * <pre>
+ * Example 1:
+ * this: from [1-100] to [501-600]
+ * input: from [10-40] to [60-90]
+ * output: from [10-40] to [560-590]
+ * Example 2 ('reverse strand exons'):
+ * this: from [1-100] to [2000-1951], [1000-951] // transcript to loci
+ * input: from [1-50] to [41-90] // CDS to transcript
+ * output: from [10-40] to [1960-1951], [1000-971] // CDS to gene loci
+ * </pre>
+ *
+ * @param map
+ * @return
+ */
+ public MapList traverse(MapList map)
+ {
+ if (map == null)
+ {
+ return null;
+ }
+
+ /*
+ * compound the ratios by this rule:
+ * A:B with M:N gives A*M:B*N
+ * reduced by greatest common divisor
+ * so 1:3 with 3:3 is 3:9 or 1:3
+ * 1:3 with 3:1 is 3:3 or 1:1
+ * 1:3 with 1:3 is 1:9
+ * 2:5 with 3:7 is 6:35
+ */
+ int outFromRatio = getFromRatio() * map.getFromRatio();
+ int outToRatio = getToRatio() * map.getToRatio();
+ int gcd = MathUtils.gcd(outFromRatio, outToRatio);
+ outFromRatio /= gcd;
+ outToRatio /= gcd;
+
+ List<int[]> toRanges = new ArrayList<>();
+ for (int[] range : getToRanges())
+ {
+ int fromLength = Math.abs(range[1] - range[0]) + 1;
+ int[] transferred = map.locateInTo(range[0], range[1]);
+ if (transferred == null || transferred.length % 2 != 0)
+ {
+ return null;
+ }
+
+ /*
+ * convert [start1, end1, start2, end2, ...]
+ * to [[start1, end1], [start2, end2], ...]
+ */
+ int toLength = 0;
+ for (int i = 0; i < transferred.length;)
+ {
+ toRanges.add(new int[] { transferred[i], transferred[i + 1] });
+ toLength += Math.abs(transferred[i + 1] - transferred[i]) + 1;
+ i += 2;
+ }
+
+ /*
+ * check we mapped the full range - if not, abort
+ */
+ if (fromLength * map.getToRatio() != toLength * map.getFromRatio())
+ {
+ return null;
+ }
+ }
+
+ return new MapList(getFromRanges(), toRanges, outFromRatio, outToRatio);
+ }
+
+ /**
+ * Answers true if the mapping is from one contiguous range to another, else
+ * false
+ *
+ * @return
+ */
+ public boolean isContiguous()
+ {
+ return fromShifts.size() == 1 && toShifts.size() == 1;
+ }
+
+ /**
+ * Returns the [start, end, start, end, ...] ranges in the 'from' range that
+ * map to positions between {@code start} and {@code end} in the 'to' range.
+ * Returns null if no mapped positions are found in start-end.
+ *
+ * @param start
+ * @param end
+ * @return
+ */
+ public int[] locateInFrom2(int start, int end)
+ {
+ List<int[]> ranges = mapBetween(start, end, toShifts, fromShifts,
+ toRatio, fromRatio);
+
+ // TODO: or just return the List and adjust calling code to match
+ return ranges.isEmpty() ? null : MappingUtils.rangeListToArray(ranges);
+ }
+
+ /**
+ * Returns the [start, end, start, end, ...] ranges in the 'to' range that map
+ * to the given start-end in the 'from' range. Returns null if either
+ * {@code start} or {@code end} is not a mapped 'from' range position.
+ *
+ * @param start
+ * @param end
+ * @return
+ */
+ public int[] locateInTo2(int start, int end)
+ {
+ List<int[]> ranges = mapBetween(start, end, fromShifts, toShifts,
+ fromRatio, toRatio);
+
+ return ranges.isEmpty() ? null : MappingUtils.rangeListToArray(ranges);
+ }
+
+ /**
+ * A helper method for navigating the mapping. Returns a (possibly empty) list
+ * of [start-end] positions in {@code ranges2} that map to positions in
+ * {@code ranges1} between {@code start} and {@code end}.
+ *
+ * @param start
+ * @param end
+ * @param ranges1
+ * @param ranges2
+ * @param wordLength1
+ * @param wordLength2
+ * @return
+ */
+ final static List<int[]> mapBetween(int start, int end,
+ List<int[]> ranges1, List<int[]> ranges2, int wordLength1,
+ int wordLength2)
+ {
+ if (end < start)
+ {
+ int tmp = end;
+ end = start;
+ start = tmp;
+ }
+
+ /*
+ * first traverse ranges1 and record count of mapped positions
+ * to any that overlap start-end
+ */
+ List<int[]> overlaps = findOverlapPositions(ranges1, start, end);
+ if (overlaps.isEmpty())
+ {
+ return overlaps;
+ }
+
+ /*
+ * convert positions to equivalent 'word' positions in ranges2
+ */
+ mapWords(overlaps, wordLength1, wordLength2);
+
+ /*
+ * walk ranges2 and record the values found at
+ * the offsets in 'overlaps'
+ */
+ List<int[]> mapped = new ArrayList<>();
+ final int s1 = overlaps.size();
+ final int s2 = ranges2.size();
+ int ranges2Index = 0;
+
+ /*
+ * count of mapped positions preceding ranges2[ranges2Index]
+ */
+ int traversed = 0;
+
+ /*
+ * for each [from-to] range in overlaps:
+ * - walk (what remains of) ranges2
+ * - record the values at offsets [from-to]
+ * - stop when past 'to' positions (or at end of ranges2)
+ */
+ for (int i = 0; i < s1; i++)
+ {
+ int[] overlap = overlaps.get(i);
+ final int toAdd = overlap[1] - overlap[0] + 1;
+ int added = 0; // how much of overlap has been 'found'
+ for (; added < toAdd && ranges2Index < s2; ranges2Index++)
+ {
+ int[] range2 = ranges2.get(ranges2Index);
+ int rangeStart = range2[0];
+ int rangeEnd = range2[1];
+ boolean reverseStrand = range2[1] < range2[0];
+ int rangeLength = Math.abs(rangeEnd - rangeStart) + 1;
+ if (traversed + rangeLength <= overlap[0])
+ {
+ /*
+ * precedes overlap - keep looking
+ */
+ traversed += rangeLength;
+ continue;
+ }
+ int overlapStart = overlap[0] - traversed;
+ int overlapEnd = Math.min(overlapStart + toAdd - added - 1,
+ rangeLength - 1);
+ int mappedFrom = range2[0] + (reverseStrand ? - overlapStart : overlapStart);
+ int mappedTo = range2[0] + (reverseStrand ? - overlapEnd : overlapEnd);
+ mapped.add(new int[] { mappedFrom, mappedTo });
+ int found = overlapEnd - overlapStart + 1;
+ added += found;
+ overlap[0] += found;
+ traversed += rangeLength;
+ }
+ }
+
+ return mapped;
+ }
+
+ /**
+ * Converts the start-end positions (counted from zero) in the {@code ranges}
+ * list from one word length to another. Start-end positions are expanded if
+ * necessary to cover a whole word of length {@code wordLength1}. Positions
+ * are then divided by {@code wordLength1} and multiplied by
+ * {@code wordLength2} to give equivalent mapped words.
+ * <p>
+ * Put simply, this converts peptide residue positions to the corresponding
+ * codon ranges, and codons - including partial codons - to the corresponding
+ * peptide positions; for example
+ *
+ * <pre>
+ * [1, 10] with word lengths 3:1 converts (as if bases [0-11]) to [1, 4]
+ * </pre>
+ *
+ * @param ranges
+ * @param wordLength1
+ * @param wordLength2
+ * @return
+ */
+ final static void mapWords(List<int[]> ranges, int wordLength1,
+ int wordLength2)
+ {
+ if (wordLength1 == 1 && wordLength2 == 1)
+ {
+ return; // nothing to do here
+ }
+ int s = ranges.size();
+ for (int i = 0; i < s; i++)
+ {
+ int[] range = ranges.get(i);
+
+ /*
+ * expand range start to the start of a word,
+ * and convert to wordLength2
+ */
+ range[0] -= range[0] % wordLength1;
+ range[0] = range[0] / wordLength1 * wordLength2;
+
+ /*
+ * similar calculation for range end, adding
+ * (wordLength2 - 1) for end of mapped word
+ */
+ range[1] -= range[1] % wordLength1;
+ range[1] = range[1] / wordLength1 * wordLength2;
+ range[1] += wordLength2 - 1;
+ }
+ }
+
+ /**
+ * Helper method that returns a (possibly empty) list of offsets in
+ * {@code ranges} to subranges that overlap {@code start-end} (where start <=
+ * end}. The list returned holds counts of the number of positions traversed
+ * (exclusive) to reach the overlapping positions, not the overlapping values.
+ * Returns null if there are no overlaps.
+ *
+ * @param ranges
+ * @param start
+ * @param end
+ * @return
+ */
+ final static List<int[]> findOverlapPositions(List<int[]> ranges,
+ int start, int end)
+ {
+ List<int[]> positions = new ArrayList<>();
+ int pos = 0;
+ int s = ranges.size();
+ for (int i = 0; i < s; i++)
+ {
+ int[] range = ranges.get(i);
+ addOverlap(positions, pos, range, start, end);
+ pos += 1 + Math.abs(range[1] - range[0]);
+ }
+ return positions;
+ }
+
+ /**
+ * A helper method that checks whether {@code range} overlaps
+ * {@code start-end}, and if so adds the offset of the overlap in
+ * {@code range}, plus {@code pos}, to {@code positions}.
+ *
+ * @param positions
+ * a list of map offsets to add to
+ * @param pos
+ * the number of mapped positions already visited
+ * @param range
+ * a from-to range (may be forward or reverse)
+ * @param start
+ * position to test for overlap in range
+ * @param end
+ * position to test for overlap in range
+ * @return
+ */
+ final static void addOverlap(List<int[]> positions, int pos, int[] range,
+ int start, int end)
+ {
+ if (range[1] >= range[0])
+ {
+ /*
+ * forward direction range
+ */
+ if (start <= range[1] && end >= range[0])
+ {
+ /*
+ * overlap
+ */
+ int overlapStart = Math.max(start, range[0]);
+ int overlapStartOffset = pos + overlapStart - range[0];
+ int overlapEnd = Math.min(end, range[1]);
+ int overlapEndOffset = pos + overlapEnd - range[0];
+ int[] lastOverlap = positions.isEmpty() ? null
+ : positions.get(positions.size() - 1);
+ if (lastOverlap != null && overlapStartOffset == lastOverlap[1] + 1)
+ {
+ /*
+ * just extending the last overlap range
+ */
+ lastOverlap[1] = overlapEndOffset;
+ }
+ else
+ {
+ /*
+ * add a new (discontiguous) overlap range
+ */
+ positions.add(new int[] { overlapStartOffset, overlapEndOffset });
+ }
+ }
+ }
+ else
+ {
+ /*
+ * reverse direction range
+ */
+ if (start <= range[0] && end >= range[1])
+ {
+ /*
+ * overlap
+ */
+ int overlapStart = Math.max(start, range[1]);
+ int overlapEnd = Math.min(end, range[0]);
+ positions
+ .add(new int[]
+ { pos + range[0] - overlapEnd,
+ pos + range[0] - overlapStart });
+ }
+ }
+ }
}