This program finds local alignments between query sequences, and reference sequences that have been prepared using lastdb. You can use it like this:

lastdb humanDb humanChromosome*.fasta
lastal humanDb dna*.fasta > myalns.maf

The lastdb command reads files called humanChromosome*.fasta and writes several files whose names begin with humanDb. The lastal command reads files called dna*.fasta, compares them to humanDb, and writes alignments to a file called myalns.maf.

You can also pipe query sequences into lastal, for example:

zcat seqs.fasta.gz | lastal humanDb > myalns.maf

Steps in lastal

  1. Find initial matches. For each possible start position in the query: find the shortest match with length ≥ l that either occurs ≤ m times in the reference, or has length L.

  2. Extend a gapless alignment from each initial match, and keep those with score ≥ d.

  3. Define cores: find the longest run of identical matches in each gapless alignment.

  4. Extend a gapped alignment from either side of each core, and keep those with score ≥ e.

  5. Non-redundantize the alignments: if several alignments share an endpoint (same coordinates in both sequences), remove all but one highest-scoring one.

  6. Estimate the ambiguity of each aligned column (OFF by default).

  7. Redo the alignments to minimize column ambiguity, using either gamma-centroid or LAMA (OFF by default).


Cosmetic options

-h, --help Show all options and their default settings, and exit.
-V, --version Show version information, and exit.
-v Be verbose: write messages about what lastal is doing.

Choose the output format. The NAME is not case-sensitive.

MAF format looks like this:

a score=15 EG2=4.7e+04 E=2.6e-05
s seqA 2 21 +     25 TTTGGGAGTTGAAGGTT--GCCC

Lines starting with "s" contain: the sequence name, the start coordinate of the alignment, the number of sequence letters spanned by the alignment, the strand, the sequence length, and the aligned letters. The start coordinates are zero-based. If the strand is "-", the start coordinate is in the reverse strand.

The same alignment in TAB format looks like this:

15 chr3 9 23 + 939557 seqA 2 21 + 25 17,2:0,4 EG2=4.7e+04 E=2.6e-05

The "17,2:0,4" shows the sizes and offsets of gapless blocks in the alignment. In this case, we have a block of size 17, then an offset of size 2 in the upper sequence and 0 in the lower sequence, then a block of size 4.

The same alignment in BlastTab format looks like this:

seqA chr3 86.96 23 1 1 3 23 10 32 2.6e-05 44.3

The fields are: query name, reference name, percent identity, alignment length, mismatches, gap opens, query start, query end, reference start, reference end, E-value, bit score. The start coordinates are one-based. Warning: this is a lossy format, because it does not show gap positions. Warning: the other LAST programs cannot read this format. Warning: "bit score" is not the same as "score".

BlastTab+ format is the same as BlastTab, with 2 extra columns at the end: length of query sequence and length of reference sequence. More columns might be added in future.

For backwards compatibility, a NAME of 0 means TAB and 1 means MAF.

E-value options

-D LENGTH Report alignments that are expected by chance at most once per LENGTH query letters. This option only affects the default value of -E, so if you specify -E then -D has no effect.
-E THRESHOLD Maximum EG2 (expected alignments per square giga). This option only affects the default value of -e, so if you specify -e then -E has no effect.

Score options

-r SCORE Match score.
-q COST Mismatch cost.

Specify a match/mismatch score matrix. Options -r and -q will be ignored. The built-in matrices are described in last-matrices.html.

Any other NAME is assumed to be a file name. For an example of the format, see the matrix files in the data directory. Any letters that aren't in the matrix will get the lowest score in the matrix when aligned to anything. Asymmetric scores are allowed: query letters correspond to columns and reference letters correspond to rows. Other options can be specified on lines starting with "#last", but command line options override them.

-a COST Gap existence cost.
-b COST Gap extension cost. A gap of size k costs: a + (b × k).
-A COST Insertion existence cost. This refers to insertions in the query relative to the reference. If -A is not used, the insertion existence cost will equal the deletion existence cost, which is set by -a.
-B COST Insertion extension cost.
-c COST This option allows use of "generalized affine gap costs" (SF Altschul 1998, Proteins 32(1):88-96). Here, a "gap" may consist of unaligned regions of both sequences. If these unaligned regions have sizes j and k, where j ≤ k, the cost is: a + b⋅(k-j) + c⋅j. If c ≥ a + 2b (the default), it reduces to standard affine gaps.

Align DNA queries to protein reference sequences, using the specified frameshift cost. A value of 15 seems to be reasonable. (As a special case, -F0 means DNA-versus-protein alignment without frameshifts, which is faster.) The output looks like this:

a score=108

The \ indicates a forward shift by one nucleotide, and the / indicates a reverse shift by one nucleotide. The * indicates a stop codon. The same alignment in tabular format looks like this:

108 prot 2 40 + 649 dna 8 117 + 999 4,1:0,6,0:1,10,0:-1,19

The "-1" indicates the reverse frameshift.

-x DROP Maximum score drop for gapped alignments. Gapped alignments are forbidden from having any internal region with score < -DROP. This serves two purposes: accuracy (avoid spurious internal regions in alignments) and speed (the smaller the faster).
-y DROP Maximum score drop for gapless alignments.
-z DROP Maximum score drop for final gapped alignments. Setting z different from x causes lastal to extend gapless alignments twice: first with a maximum score drop of x, and then with a (presumably higher) maximum score drop of z.
-d SCORE Minimum score for gapless alignments.
-e SCORE Minimum alignment score. (If you do gapless alignment with option -j1, then -d and -e mean the same thing. If you set both, -e will prevail.)

Initial-match options


Maximum multiplicity for initial matches. Each initial match is lengthened until it occurs at most this many times in the reference.

If the reference was split into volumes by lastdb, then lastal uses one volume at a time. The maximum multiplicity then applies to each volume, not the whole reference. This is why voluming changes the results.

-l LENGTH Minimum length for initial matches. Length means the number of letters spanned by the match.
-L LENGTH Maximum length for initial matches.
-k STEP Look for initial matches starting only at every STEP-th position in each query (positions 0, STEP, 2×STEP, etc). This makes lastal faster but less sensitive.
-W SIZE Look for initial matches starting only at query positions that are "minimum" in any window of SIZE consecutive positions (see lastdb.html). By default, this parameter takes the same value as was used for lastdb -W.

Miscellaneous options

-s STRAND Specify which query strand should be used: 0 means reverse only, 1 means forward only, and 2 means both.
-S NUMBER Specify which DNA strand the score matrix applies to. This matters only for unusual matrices that lack strand symmetry (e.g. if the a:g score differs from the t:c score). 0 means that the matrix applies to the forward strand of the reference aligned to either strand of the query. 1 means that the matrix applies to either strand of the reference aligned to the forward strand of the query.
-K LIMIT Omit any alignment whose query range lies in LIMIT or more other alignments with higher score (and on the same strand). This is a useful way to get just the top few hits to each part of each query (P Berman et al. 2000, J Comput Biol 7:293-302).
-C LIMIT Before extending gapped alignments, discard any gapless alignment whose query range lies in LIMIT or more others (for the same strand and volume) with higher score-per-length. This can reduce run time and output size (MC Frith & R Kawaguchi 2015, Genome Biol 16:106).
-P THREADS Divide the work between this number of threads running in parallel. 0 means use as many threads as your computer claims it can handle simultaneously. Single query sequences are not divided between threads, so you need multiple queries per batch for this option to take effect.

Search queries in batches of at most this many bytes. If a single sequence exceeds this amount, however, it is not split. You can use suffixes K, M, and G to specify KibiBytes, MebiBytes, and GibiBytes. This option has no effect on the results (apart from their order).

If the reference was split into volumes by lastdb, then each volume will be read into memory once per query batch.


Find minimum-difference alignments, which is faster but cruder. This treats all matches the same, and minimizes the number of differences (mismatches plus gaps).

  • Any substitution score matrix will be ignored. The substitution scores are set by the match score (r) and the mismatch cost (q).

  • The gap cost parameters will be ignored. The gap existence cost will be 0 and the gap extension cost will be q + r/2.

  • The match score (r) must be an even number.

  • Any sequence quality data (e.g. fastq) will be ignored.


Type of alignment: 0 means "local alignment" and 1 means "overlap alignment". Local alignments can end anywhere in the middle or at the ends of the sequences. Overlap alignments must extend to the left until they hit the end of a sequence (either query or reference), and to the right until they hit the end of a sequence.

Warning: it's often a bad idea to use -T1. This setting does not change the maximum score drop allowed inside alignments, so if an alignment cannot be extended to the end of a sequence without exceeding this drop, it will be discarded.

-n COUNT Maximum number of gapless alignments per query position. When lastal extends gapless alignments from initial matches that start at one query position, if it gets COUNT successful extensions, it skips any remaining initial matches starting at that position.
-N COUNT Stop after finding COUNT alignments per query strand. This makes lastal faster: it quits gapless and gapped extensions as soon as it finds COUNT alignments with score ≥ e.

Specify lowercase-marking of repeats, by two digits (e.g. "-R 01"), with the following meanings.

First digit:

  1. Convert the input sequences to uppercase while reading them.

  2. Keep any lowercase in the input sequences.

Second digit:

  1. Do not check for simple repeats.

  2. Convert simple repeats (e.g. cacacacacacacacac) to lowercase.

  3. Convert simple repeats, within AT-rich DNA, to lowercase.

Details: Tantan is applied separately to forward and reverse strands. For DNA-versus-protein alignment (option -F), it is applied to the DNA after translation, at the protein level.


Specify treatment of lowercase letters when extending alignments:

  1. Mask them for neither gapless nor gapped extensions.

  2. Mask them for gapless but not gapped extensions.

  3. Mask them for gapless but not gapped extensions, and then discard alignments that lack any segment with score ≥ e when lowercase is masked.

  4. Mask them for gapless and gapped extensions.

"Mask" means change their match/mismatch scores to min(unmasked score, 0). This option does not affect treatment of lowercase for initial matches.


This option is a kludge to avoid catastrophic time and memory usage when self-comparing a large sequence. If the sequence contains a tandem repeat, we may get a gapless alignment that is slightly offset from the main self-alignment. In that case, the gapped extension might "discover" the main self-alignment and extend over the entire length of the sequence.

To avoid this problem, gapped alignments are not triggered from any gapless alignment that:

  • is contained, in both sequences, in the "core" of another alignment

  • has start coordinates offset by DISTANCE or less relative to this core

Use -w0 to turn this off.

-G FILE Use an alternative genetic code in the specified file. For an example of the format, see vertebrateMito.gc in the examples directory. By default, the standard genetic code is used. This option has no effect unless DNA-versus-protein alignment is selected with option -F.
-t TEMPERATURE Parameter for converting between scores and likelihood ratios. This affects the column ambiguity estimates. A score is converted to a likelihood ratio by this formula: exp(score / TEMPERATURE). The default value is 1/lambda, where lambda is the scale factor of the scoring matrix, which is calculated by the method of Yu and Altschul (YK Yu et al. 2003, PNAS 100(26):15688-93).

This option affects gamma-centroid and LAMA alignment only.

Gamma-centroid alignments minimize the ambiguity of paired letters. In fact, this method aligns letters whose column error probability is less than GAMMA/(GAMMA+1). When GAMMA is low, it aligns confidently-paired letters only, so there tend to be many unaligned letters. When GAMMA is high, it aligns letters more liberally.

LAMA (Local Alignment Metric Accuracy) alignments minimize the ambiguity of columns (both paired letters and gap columns). When GAMMA is low, this method produces shorter alignments with more-confident columns, and when GAMMA is high it produces longer alignments including less-confident columns.

In summary: to get the most accurately paired letters, use gamma-centroid. To get accurately placed gaps, use LAMA.

Note that the reported alignment score is that of the ordinary gapped alignment before realigning with gamma-centroid or LAMA.


Output type: 0 means counts of initial matches (of all lengths); 1 means gapless alignments; 2 means gapped alignments before non-redundantization; 3 means gapped alignments after non-redundantization; 4 means alignments with ambiguity estimates; 5 means gamma-centroid alignments; 6 means LAMA alignments; 7 means alignments with expected counts.

If you use -j0, lastal will count the number of initial matches, per length, per query sequence. Options -l and -L will set the minimum and maximum lengths, and -m will be ignored. If you compare a large sequence to itself with -j0, it's wise to set option -L.

If you use j>3, each alignment will get a "fullScore" (also known as "forward score" or "sum-of-paths score"). This is like the score, but it takes into account alternative alignments.

If you use -j7, lastal will print an extra MAF line starting with "c" for each alignment. The first 16 numbers on this line are the expected counts of matches and mismatches: first the count of reference As aligned to query As, then the count of reference As aligned to query Cs, and so on. For proteins there will be 400 such numbers. The final ten numbers are expected counts related to gaps. They are:

  • The count of matches plus mismatches. (This may exceed the total of the preceding numbers, if the sequences have non-ACGT letters.)

  • The count of deleted letters.

  • The count of inserted letters.

  • The count of delete opens (= count of delete closes).

  • The count of insert opens (= count of insert closes).

  • The count of adjacent pairs of insertions and deletions.

The final four numbers are always zero, unless you use generalized affine gap costs. They are:

  • The count of unaligned letter pairs.

  • The count of unaligned letter pair opens (= count of closes).

  • The count of adjacent pairs of deletions and unaligned letter pairs.

  • The count of adjacent pairs of insertions and unaligned letter pairs.


This option allows lastal to use sequence quality scores, or PSSMs, for the queries. 0 means read queries in fasta format (without quality scores); 1 means fastq-sanger format; 2 means fastq-solexa format; 3 means fastq-illumina format; 4 means prb format; 5 means read PSSMs. (Warning: Illumina data is not necessarily in fastq-illumina format; it is often in fastq-sanger format.)

The fastq formats look like this:


The "+" may optionally be followed by a name (ignored), and the sequence and quality codes are allowed to wrap onto more than one line. For fastq-sanger, the quality scores are obtained by subtracting 33 from the ASCII values of the characters below the "+". For fastq-solexa and fastq-illumina, they are obtained by subtracting 64.

prb format stores four quality scores (A, C, G, T) per position, with one sequence per line, like this:

-40  40 -40 -40      -12   1 -12  -3      -10  10 -40 -40

Since prb does not store sequence names, lastal uses the line number (starting from 1) as the name.

In fastq-sanger and fastq-illumina format, the quality scores are related to error probabilities like this: qScore = -10⋅log10[p]. In fastq-solexa and prb, however, qScore = -10⋅log10[p/(1-p)]. In lastal's MAF output, the quality scores are written on lines starting with "q". For fastq, they are written with the same encoding as the input. For prb, they are written in the fastq-solexa (ASCII-64) encoding.

Finally, PSSM means "position-specific scoring matrix". The format is:

     A  R  N  D  C  Q  E  G  H  I  L  K  M  F  P  S  T  W  Y  V
1 M -2 -2 -3 -4 -2 -1 -3 -3 -2  1  2 -2  8 -1 -3 -2 -1 -2 -2  0
2 S  0 -2  0  1  3 -1 -1 -1 -2 -3 -3 -1 -2 -3 -2  5  0 -4 -3 -2
3 D -1 -2  0  7 -4 -1  1 -2 -2 -4 -4 -2 -4 -4 -2 -1 -2 -5 -4 -4

The sequence appears in the second column, and columns 3 onwards contain the position-specific scores. Any letters not specified by any column will get the lowest score in each row. This format is a simplified version of PSI-BLAST's ASCII format: the non-simplified version is allowed too.

Warning: lastal cannot directly calculate E-values for PSSMs. The E-values (and the default value of -y) are determined by the otherwise-unused match and mismatch scores (options -r -q and -p). There is evidence these E-values will be accurate if the PSSM is "constructed to the same scale" as the match/mismatch scores (SF Altschul et al. 1997, NAR 25(17):3389-402).

Parallel processes and memory sharing

If you run several lastal commands (i.e. processes) at the same time on the same computer, using the same set of reference files prepared by lastdb, then they will share memory for the reference files.

Multiple volumes

If lastdb creates multiple volumes:

lastdb hugeDb huge.fasta

You can either run lastal on the whole thing:

lastal hugeDb queries.fasta > myalns.maf

Or on one volume at a time:

lastal hugeDb0 queries.fasta > myalns0.maf
lastal hugeDb1 queries.fasta > myalns1.maf
lastal hugeDb2 queries.fasta > myalns2.maf

The former method reads the queries in large batches, and aligns each batch to one volume at a time. If you run several processes in parallel, they will not necessarily use the same volume at the same time.

Therefore, with parallel processes, you should either ensure you have enough memory to hold several volumes simultaneously, or run lastal on one volume at a time. An efficient scheme is to use a different computer for each volume.


lastal8 has identical usage to lastal, and is used with lastdb8. lastal cannot read the output of lastdb8, and lastal8 cannot read the output of lastdb.