FAQs

LexicMap is mainly designed for sequence alignment with a small number of queries (gene/plasmid/virus/phage sequences) longer than 200 bp by default. However, short queries can also be aligned.

If you just want to search long (>1kb) queries for highy similar (>95%) targets, you can build an index with a bigger -D/--seed-max-desert (200 by default), e.g.,

--seed-max-desert 450 --seed-in-desert-dist 150

Bigger values decrease the search sensitivity for distant targets, speed up the indexing speed, decrease the indexing memory occupation and decrease the index size. While the alignment speed is almost not affected.

Yes. LexicMap mainly supports small genomes including prokaryotic, viral, and plasmid genomes. Fungi can also be supported, just remember to increase the value of -g/--max-genome when running lexicmap index, which is used to skip genomes larger than 15Mb by default.

  -g, --max-genome int            ► Maximum genome size. Extremely large genomes (e.g., non-isolate
                                  assemblies from Genbank) will be skipped. (default 15000000)

Maximum genome size is about 268 Mb (268,435,456). More precisely:

$total_bases + ($num_contigs - 1) * 1000 <= 268,435,456

as we concatenate contigs with 1000-bp intervals of N’s to reduce the sequence scale to index.

For big and complex genomes, like the human genome (chr1 is ~248 Mb) which has many repetitive sequences, LexicMap would be slow to align.

• For index building. See details hardware requirement.
• For seaching. See details hardware requirement.

Use lexicmap utils remerge (available since v0.5.0), which reruns the merging step for an unfinished index.

When to use this command?

• Only one thread is used for merging indexes, which happens when there are a lot (>200 batches) of batches ($inpu_files / --batch-size) and the value of --max-open-files is not big enough.
• The Slurm/PBS job time limit is almost reached and the merging step won’t be finished before that.
• Disk quota is reached in the merging step.

So you can stop the indexing command by press Ctrl + C (make sure it is in the merging step, see example below), and run lexicmap utils remerge -d index.lmi, where index.lmi is the output index directory in lexicmap index.

Optionally, you might set bigger values of flag --max-open-files and -J/--seed-data-threads if you have hundreds of thousands of input genomes or have set a small batch size with -b/--batch-size. E.g.,

22:54:24.420 [INFO] merging 297 indexes...
22:54:24.455 [INFO]   [round 1]
22:54:24.455 [INFO]     batch 1/1, merging 297 indexes to xxx.lmi.tmp/r1_b1 with 1 threads...

There’s only one thread was used for seed data merging, it would take a long time. So we can set a larger --max-open-files, e.g., 4096, and it would allow 4096 / (297+2) = 13.7 threads for merging, let’s set --seed-data-threads 12.

# specify the maximum open files per process
ulimit -n 4096

lexicmap utils remerge -d index.lmi --max-open-files 4096 --seed-data-threads 12

Yes, lexicmap search has a flag

  -a, --all                            ► Output more columns, e.g., matched sequences. Use this if you
                                       want to output blast-style format with "lexicmap utils 2blast".

to output CIGAR string, aligned query and subject sequences.

18. cigar,    CIGAR string of the alignment                       (optional with -a/--all)
19. qseq,     Aligned part of query sequence.                     (optional with -a/--all)
20. sseq,     Aligned part of subject sequence.                   (optional with -a/--all)
21. align,    Alignment text ("|" and " ") between qseq and sseq. (optional with -a/--all)

An example:

# Extracting similar sequences for a query gene.

# search matches with query coverage >= 90%
lexicmap search -d gtdb_complete.lmi/ b.gene_E_faecalis_SecY.fasta -o results.tsv \
    --min-qcov-per-hsp 90 --all

# extract matched sequences as FASTA format
sed 1d results.tsv | awk -F'\t' '{print ">"$5":"$14"-"$15":"$16"\n"$20;}' \
    | seqkit seq -g > results.fasta

seqkit head -n 1 results.fasta | head -n 3
>NZ_JALSCK010000007.1:39224-40522:-
TTGTTCAAGCTATTAAAGAACGCCTTTAAAGTCAAAGACATTAGATCAAAAATCTTATTT
ACAGTTTTAATCTTGTTTGTATTTCGCCTAGGTGCGCACATTACTGTGCCCGGGGTGAAT

And lexicmap util 2blast can help to convert the tabular format to Blast-style format, see examples.

lexicmap utils subseq can extract subsequencess via genome ID, sequence ID and positions. So you can use these information from the search result and expand the region positions to extract flanking sequences.

It happens if there are some degenerate bases (e.g., N) in the query sequence. In the indexing step, all degenerate bases are converted to their lexicographic first bases. E.g., N is converted to A. While for the query sequences, we don’t convert them.

LexicMap is mainly designed for sequence alignment with a small number of queries against a database with a huge number (up to 17 million) of genomes. There are some ways to improve the search speed of lexicmap search.

• Increasing the concurrency number

  • Make sure that the value of -j/--threads (default: all available CPUs) is ≥ than the number of seed chunk file (default: all available CPUs in the indexing step), which can be found in info.toml file, e.g,

    # Seeds (k-mer-value data) files
    chunks = 48
    

  • Increasing the value of --max-open-files (default 1024). You might also need to change the open files limit.

  • (If you have many queries) Increase the value of -J/--max-query-conc (default 12), it will increase the memory.

• Loading the entire seed data into memoy (It’s unnecessary if the index is stored in SSD)
  • Setting -w/--load-whole-seeds to load the whole seed data into memory for faster search. For example, for ~85,000 GTDB representative genomes, the memory would be ~260 GB with default parameters.
• Returning less results
  • Setting -n/--top-n-genomes to keep top N genome matches for a query (0 for all) in chaining phase. For queries with a large number of genome hits, a resonable value such as 1000 would reduce the computation time.
• Sacrificing accuracy
  • Setting --pseudo-align to only perform pseudo alignment, which is slightly faster and uses less memory. It can be used in searching with long and divergent query sequences like nanopore long-reads.

Click to read more detail of the usage.