DCD-Database:
- directory called joined_sort_clust_split
- contains 3786 files
- 3 Columns, Tab seperated: 1. Sequence ID; 2. Sequence; 3. Cluster ID
- sorted lexicographically, File name is first Cluster ID in this file.
search_seqs_to_ffformat.sh:
Takes a centroid name and a integer to infer tmp file.
Searches for all sequences which are part of the cluster represented by the centroid.
Further caluclates MSA for this cluster with FAMSA2.
run_random_search_seqs.sh:
Takes path to a file with list of centroids and integer to infer tmp file.
Calls search_seqs_to_ffformat.sh and builds ffindex formated file for the clusters.
rand_msa_evaluation.sh:
Counting Number of Sequences in MSAs created by AF2
Commands to create HMM-Profiles:
Add hhconsensus to each cluster in ffindex file (can take really long):
/PATH_TO/ffindex_apply DCDB_msa.ff{data,index} -i DCDB_a3m.ffindex -d DCDB_a3m.ffdata -- /PATH_TO/hhconsensus -M 50 -maxseq 150000 -maxres 115535 -i stdin -oa3m stdout -v 0
Calculate HMM-Profiles for each cluster in ffindex file:
/PATH_TO/ffindex_apply DCDB_a3m.ffdata DCDB_a3m.ffindex -i DCDB_hhm.ffindex -d DCDB_hhm.ffdata -- /PATH_TO/hhmake -i stdin -o stdout -v 0
Call cstranslate to add pseudo counts:
/PATH_TO/cstranslate -f -x 0.3 -c 12 -I a3m -i DCDB_a3m -o DCDB_cs219
Combine two FFindex files with:
/PATH_TO/ffindex_build -s -a DCDB_1_a3m.ff{data,index} -d DCDB_2_a3m.ffdata -i ../random_corr/DCDB_2_a3m.ffindex
Sort Cluster by Size:
sort -k3 -n -r DCDB_cs219.ffindex | cut -f1 > sorting.dat
Reorder FFindex Database by sorting.dat:
/PATH_TO/ffindex_order sorting.dat /PATH_TO/DCDB_hhm.ff{data,index} /PATH_TO/DCDB_hhm_ordered.ff{data,index}\
/PATH_TO/ffindex_order sorting.dat /PATH_TO/DCDB_a3m.ff{data,index} /PATH_TO/DCDB_a3m_ordered.ff{data,index}
Commands for Alignment Tools:
Jackhmmer query against Centroids:
/PATH_TO/jackhmmer --cpu 32 --tblout alignment.out -A alignment.sto --acc --noali --F1 0.0005 --F2 0.00005 --F3 0.0000005 --incE 0.0001 -E 0.0001 -N 1 sequence.fasta /path_to/centroids_smaller_than_100k.dedup.faa
Jackhmmer against Mgnify-Database to reduce expensive alignment runs:
/PATH_TO/jackhmmer --cpu 32 --tblout alignment.out --noali --notextw --incE 0.1 -E 0.1 sequence.fasta /PATH_TO/mgy_clusters_2022_05.fa
HHblits against BFD and Uniref30 to reduce expensive alignment runs (NOTE: a3m and tsv seem to report hits different, AF2 uses a3m):
/PATH_TO/hhblits -i sequence.fasta -cpu 18 -oa3m sequence.a3m -blasttab sequence.tsv -n 3 -e 0.001 -maxseq 1000000 -realign_max 100000 -maxfilt 100000 -min_prefilter_hits 1000 -d /PATH_TO/bfd_metaclust_clu_complete_id30_c90_final_seq.sorted_opt -d /PATH_TO/UniRef30_2021_03
Additional Information:
- AlphaFold2:
- AF2 parameters:--run_msa_and_templates_only --nouse_gpu_relax
- maximum template date set to 2022-12-31
- HHblits: set Parameter FFINDEX_MAX_ENTRY_NAME_LENGTH to 128 (in code)