ndaniels / mica Goto Github PK

Thanks for making this tool but I have trouble installing it. It gives following error after typing this command:

go get github.com/ndaniels/MICA/...

import cycle not allowed
package github.com/ndaniels/MICA
imports bytes
imports errors
imports runtime
imports unsafe
imports runtime

Hi,
I'm getting an error
Error processing queries: Error converting diamond output to blast tabular: %!s(MISSING)

To test mica's blastx function I created a test set using some contigs I had lying around. I used prodigal to predict the proteins from these contigs and compressed them to form the database. I then used the first contig as the query. (see attached).

cmds to duplicate error:

mica-compress --match-seq-id-threshold 60 --ext-seed-size 0 --ext-seq-id-threshold 50 --max-seeds 20 -p 40 testdb short.faa
mica-xsearch --dmnd-fine=result.txt testdb test.fasta

If I run blastx using the same query("test.fasta") and protein file("short.faa") for the database, I get the expected query hitting itself.

makeblastdb -dbtype prot -in short.faa
blastx -query test.fasta -db short.faa -outfmt '6 qseqid sseqid pident length qlen slen evalue bitscore' -out test.blastx  -max_target_seqs 1

Scaffold2.1.1-9897.test Scaffold2.1.1-9897_1_3 100.00 389 3600 389 0.0 780

Any advice?
Thanks

test.zip

Hi,
I need to compress the uniref90 database, which contains 35218174 sequences。 After running 50 hours, finally exit with "out of memory" ERROR。

How could I speed up this step?

How to avoid the ERROR ?

thanks !

Hello! Could you create a new database for mica with a more up to date version of nr? The 2015 version is a bit out of date at this point. Thanks!

In the README under the compatibility note section, you link to the version of DIAMOND used for the paper but the link is broken and returns "connection refused".

Hi there,
The README file says that the easiest way to use the program is to download the precompiled binaries. However the link to the linux 64 package is currently broken (http://giant.csail.mit.edu/gems/mica-linux-amd64.tgz). Any chance to get it fixed?
thanks!

I'm getting the following error while using MICA

Error processing queries: Could not decompress coarse sequence 16859848 (16, 43): Could not read compressed sequence: Could not scan coarse sequence residues of 16859848 from coarse-file: input does not match format

This comes from the following segment in coarse.go

n, err := fmt.Fscanf(coarsedb.FileFasta, ">%d\n%s\n", &corSeqId, &residues)

if err != nil {
        return nil, fmt.Errorf("Could not scan coarse sequence residues of %d from coarse-file: %s", id, err)
}

However the relevant lines in coarse.fasta appears to be fine:

>16859848
MKLTEKKKEEWEKLLRDTALIGPIDTIQEHTQGDLWEFGSQTRGNYFFSTEKFVFVGGLGGLTNFSVPYKNIKELKLCNIGGLIPIIPTGIKVTYTDENGKSVKKKCSVMKRKDWLAYLQEKSGIS

The database available on gems.csail.mit.edu at the time of writing is improperly formatted and will cause MICA to fail.

The headers in the coarse.fasta file are formatted as '> seqname' which diamond interprets as a blank name.

The current version of mica addresses this issue but the database has not yet been updated.

This can be corrected by running the following command from within the database directory:

mv coarse.fasta coarse.fasta.bak; cat coarse.fasta.bak | sed 's/> />/g' > coarse.fasta

mica-psisearch neglects to set the num_threads parameter for psi_blast
pull-request #9 fixes this and significantly improves mica-psisearch's performance on multithreaded machines