Dear Maxime,

I notice that when I use kaammer -server with 64 threads for searching (refseq-bacteria database), the kaamer -search gradually takes more and more memory until the maximum of the machine (512G). I am wondering how Mac memory I should assign to the searching process on a cluster to make sure that it is the most efficient and I do not waste any memory.

Thank you,

Jianshu

I have a question regarding the following sentences in the manuscript

The fixed k-mer size of 7 was chosen to fit on 4 bytes and keep a manageable database size while offering good specificity over protein targets.

The first KV store (k-mer store) keeps the association of every k-mer (key) with a hash value (key length: 8 bytes) that is the entry to the combination store. 

Is this saying that 7-mers are first hashed into 4-bytes and then used to associate with an 8-bytes hash value?

Dear Maxime,

This is Jianshu, a bioinformatics PhD student at Georgia Tech. I am reaching out to you to discuss the possibility of applying the protein database structure you proposed in the kaamer software to average amino acid identity (AAI) calculation in very large data collection (e.g., 500,000 genome or more). See the definition for AAI here: http://enve-omics.ce.gatech.edu/aai/

AAI is also widely applied in taxonomy identification of bacterial genomes if we have a large collection of bacterial genome database to compare against with. Finding the best hit in a database is quite fast after trying kaamer, but how can we do it for each genome given a identity threshold (if a protein in the query is large than ~30% identity with a protein in a database for a genome, then we can use this protein for AAI between the genome in the database and the genome the query protein belongs to). What is the smallest identity for kaamer search to find a hit in the database? I will be very happy to talk.

Thanks,

Jianshu

Hi @zorino,

Thanks for creating KAAmer and making it public!

I was planning to use KAAmer to search for very small peptides (down to 8 AA), but can't get it to find control peptides smaller than 13AA.

I use the following example data:

>Type1Collagen-1-20
MFSFVDLRLLLLLAATALLT
>Type1Collagen-1-13
MFSFVDLRLLLLL
>Type1Collagen-1-12
MFSFVDLRLLLL

and search against the human proteome (https://www.uniprot.org/proteomes/UP000005640).

Type1Collagen-1-20 and Type1Collagen-1-13 are found with full length, 100% matches. Type1Collagen-1-12 (and any smaller sequences) are not found however. My understanding was that KAAmer should be able to find any peptide with length of at least 7, (because that's your fixed kmer size) as long as I set mink to 1, which I did. Am I doing this wrongly?

Many thanks,
Andreas

Hello Zerino,

I followed the install guide and run go install ./... in the directory in Go the newest version for MacOS M1, but I got nothing out. no info after running which kaamer

Any idea.

Thanks,

Jianshu