To run pipeline for cdr1 seqs run:
bash run_cdr1.sh
To generate simulation data and run dependency search run:
cd simulatin && bash run_simulation.sh
Runs pipeline on your data
Flags:
-ffile with stopwords-iinput file with fasta seqs-ooutput csv file-ddir to save images-afile to safe alignment to
see example at run_cdr1.sh
In simulation dependent positions are generated in pairs (1 depends on 0, 2 on 3...), also we can add positions with random letters at the end
Each 2 positions (0 1, 2 3, ...) have single unique pair of dependent letters
Usage:
# generates strings of 8 letters, 2 last are random
python simulate.py \
--p_i_range 0.04 0.07 \
--p_ji_range 0.4 0.6 \
--n_seq 5000 \
--n_pairs 3 \
--n_randoms 2 \
--seed 123 \
> data.txt
Leaves unique sequences, removes stopwords and filters by sequence length
Usage:
python filter.py \
--filter_seq_file data/cdr1_filter_seq.txt \
--min_len 8 \
--max_len 8 \
< cdr1_aa_len500.txt \
> cdr1_aa_len500_filtered.txt
Runs search of dependent positions
Usage:
python process_aln.py \
--out_dir results \
--ratio_cutoff 0 \
--c_ij_cutoff 20 \
--p_j_cond_i_cutoff auto \
--p_ij_cutoff auto \
--uncommon \
< data/cdr1_clustalo.txt \
> results.csv
--uncommonflag removes item from results if dependent letter is most common at it's position--*_cutoffparams are cutoffs for different values, some cutoffs can be set toauto,automeans that cutoff will be calculated dynamically using normal distribution assumption and 3-sigma rule
Runs simulation with different p_ji values
Usage:
--n_trialsnumber of runs perp_jivalue--p_ji_rangemin/max values ofp_ji--num_stepsnumberp_jipoints fromp_ji_range--n_procnumber of processes
python delta_simulation.py \
--p_i_range 0.04 0.1 \
--n_trials 3 \
--n_seq 1000 \
--n_pairs 3 \
--n_randoms 2 \
--seed 7 \
--p_ji_range 0.05 1 \
--num_steps 77 \
--n_proc 4 \
> delta_results/delta_results.csv
Extracts some seqs with highly dependent positions
Usage:
bash extract_seqs.sh {pipeline_results.csv} {alignment_file}
Example:
bash extract_seqs.sh data/results.csv data/cdr1_clustalo.txt > data/extracted_seqs.txt
Output format:
_ before codition position ^ before dependent position
Position (5 L) depends on position (8 L) with p=0.74
--GYT^LTE_LS-
--GHT^LSE_LS-
--GYT^LTE_LP-
React
Typescript
Django
Laravel
Facebook
Microsoft
Google
Alibaba
D3
Tencent