Mercurial > repos > jose_duarte > phagedpo
changeset 14:85d70111f4b4 draft
Deleted selected files
| author | jose_duarte |
|---|---|
| date | Wed, 08 Dec 2021 10:29:01 +0000 |
| parents | f51906b109cd |
| children | 0940c145e15a |
| files | ANN4311 DPOGALAXY.py PhageDPO.xml RF5748 SVM4311 d4311_SCALER d5748_SCALER |
| diffstat | 7 files changed, 0 insertions(+), 232 deletions(-) [+] |
line wrap: on
line diff
--- a/DPOGALAXY.py Fri Nov 26 14:42:21 2021 +0000 +++ /dev/null Thu Jan 01 00:00:00 1970 +0000 @@ -1,164 +0,0 @@ -#print('Hello world') -#PS C:\Users\joseduarte\Documents\pythonfiles\phage> python pdpo_test.py -#Hello world - -class PDPOPrediction: - def __init__(self, Folder = 'location', mdl='',seq_file = 'fasta_file.fasta',ttable=11): - import pickle - import pandas as pd - from Bio import SeqIO - import os - from pathlib import Path - self.data = {} - self.df_output = None - self.seqfile = seq_file - self.__location__ = os.path.realpath(os.path.join(os.getcwd(), Folder)) - - with open(os.path.join(self.__location__,mdl), 'rb') as m: - self.model = pickle.load(m) - if mdl == 'SVM4311': - with open(os.path.join(__location__,'d4311_SCALER'),'rb') as sl: - self.scaler = pickle.load(sl) - self.name = mdl - elif mdl == 'RF5748': - with open(os.path.join(__location__,'d5748_SCALER'),'rb') as sc: - self.scaler = pickle.load(sc) - self.name = mdl - elif mdl == 'ANN4311': - with open(os.path.join(__location__,'d4311_SCALER'),'rb') as sl: - self.scaler = pickle.load(sl) - self.name = mdl - - for seq in SeqIO.parse(os.path.join(self.__location__,self.seqfile), 'fasta'): - #name_seq = seq.id - DNA_seq = seq.seq - AA_seq = DNA_seq.translate(table=ttable) - descr_seq = seq.description.replace(' ','') - self.data[descr_seq]=[DNA_seq._data,AA_seq._data] - self.df = pd.DataFrame({'ID':list(self.data.keys()), - 'DNAseq':[elem[0] for elem in self.data.values()], - 'AAseq':[elem[1] for elem in self.data.values()]}) - self.df = self.df.set_index('ID') - - def Datastructure(self): - import pandas as pd - import pickle - from Bio.SeqUtils.ProtParam import ProteinAnalysis - from propy import CTD - from propy import AAComposition - - def count_orf(orf_seq): - dic = {'DNA-A': 0, 'DNA-C': 0, 'DNA-T': 0, 'DNA-G': 0, 'DNA-GC': 0} - for letter in range(len(orf_seq)): - for k in range(0, 4): - if orf_seq[letter] in list(dic.keys())[k][-1]: - dic[list(dic.keys())[k]] += 1 - dic['DNA-GC'] = ((dic['DNA-C'] + dic['DNA-G']) / ( - dic['DNA-A'] + dic['DNA-C'] + dic['DNA-T'] + dic['DNA-G'])) * 100 - return dic - - def count_aa(aa_seq): - dic = {'G': 0, 'A': 0, 'L': 0, 'V': 0, 'I': 0, 'P': 0, 'F': 0, 'S': 0, 'T': 0, 'C': 0, - 'Y': 0, 'N': 0, 'Q': 0, 'D': 0, 'E': 0, 'R': 0, 'K': 0, 'H': 0, 'W': 0, 'M': 0} - for letter in range(len(aa_seq)): - if aa_seq[letter] in dic.keys(): - dic[aa_seq[letter]] += 1 - return dic - - def sec_st_fr(aa_seq): - from Bio.SeqUtils.ProtParam import ProteinAnalysis - st_dic = {'Helix': 0, 'Turn': 0, 'Sheet': 0} - stu = ProteinAnalysis(aa_seq).secondary_structure_fraction() - st_dic['Helix'] = stu[0] - st_dic['Turn'] = stu[1] - st_dic['Sheet'] = stu[2] - return st_dic - - self.feat={"SVM4311": ["DNA-A", "DNA-T", "DNA-G", "DNA-GC", "AA_Len", "G", "A", "S", "T", "N", "Turn", "Sheet", - "_PolarizabilityC1", "_PolarizabilityC3", "_SolventAccessibilityC1", "_SecondaryStrC1", - "_SecondaryStrC2", "_SecondaryStrC3", "_ChargeC2", "_ChargeC3", "_PolarityC1", "_NormalizedVDWVC1", - "_NormalizedVDWVC3", "_HydrophobicityC2", "_HydrophobicityC3", "_SecondaryStrT23", - "_NormalizedVDWVT13", "_PolarizabilityD1001", "_SolventAccessibilityD1001", - "_SolventAccessibilityD2001", "_SolventAccessibilityD3001", "_SecondaryStrD1025", "_ChargeD1075", - "_ChargeD2001", "_ChargeD2025", "_ChargeD3025", "_ChargeD3050", "_PolarityD1075", "_PolarityD3025", - "_NormalizedVDWVD1001", "_NormalizedVDWVD3050", "_HydrophobicityD2001", "DG", "DT", "GD"], - "RF5748": ["DNA-A", "DNA-T", "DNA-G", "DNA-GC", "AA_Len", "G", "A", "S", "T", "N", "Turn", "Sheet", - "_PolarizabilityC1", "_PolarizabilityC3", "_SecondaryStrC1", "_SecondaryStrC2", "_SecondaryStrC3", - "_ChargeC1", "_ChargeC2", "_ChargeC3", "_NormalizedVDWVC1", "_NormalizedVDWVC3", "_HydrophobicityC2", - "_HydrophobicityC3", "_SolventAccessibilityT12", "_SolventAccessibilityT13", "_SecondaryStrT23", - "_NormalizedVDWVT23", "_HydrophobicityT12", "_PolarizabilityD1001", "_SolventAccessibilityD1001", - "_SolventAccessibilityD2001", "_SolventAccessibilityD3001", "_SecondaryStrD1001", - "_SecondaryStrD1025", "_ChargeD1025", "_ChargeD1075", "_ChargeD2001", "_ChargeD2025", "_ChargeD3025", - "_ChargeD3050", "_PolarityD1001", "_PolarityD1050", "_PolarityD1075", "_PolarityD3025", - "_NormalizedVDWVD1001", "_NormalizedVDWVD3001", "_HydrophobicityD1001", "_HydrophobicityD2001", "NG", - "DG", "DT", "GD", "GT"], - "ANN4311": ["DNA-A", "DNA-T", "DNA-G", "DNA-GC", "AA_Len", "G", "A", "S", "T", "N", "Turn", "Sheet", - "_PolarizabilityC1", "_PolarizabilityC3", "_SolventAccessibilityC1", "_SecondaryStrC1", - "_SecondaryStrC2", "_SecondaryStrC3", "_ChargeC2", "_ChargeC3", "_PolarityC1", "_NormalizedVDWVC1", - "_NormalizedVDWVC3", "_HydrophobicityC2", "_HydrophobicityC3", "_SecondaryStrT23", - "_NormalizedVDWVT13", "_PolarizabilityD1001", "_SolventAccessibilityD1001", - "_SolventAccessibilityD2001", "_SolventAccessibilityD3001", "_SecondaryStrD1025", "_ChargeD1075", - "_ChargeD2001", "_ChargeD2025", "_ChargeD3025", "_ChargeD3050", "_PolarityD1075", "_PolarityD3025", - "_NormalizedVDWVD1001", "_NormalizedVDWVD3050", "_HydrophobicityD2001", "DG", "DT", "GD"]} - - self.df_output = self.df.copy() - self.df_output.drop(['DNAseq','AAseq'],axis=1,inplace=True) - dna_feat = {} - aa_len = {} - aroma_dic = {} - iso_dic = {} - aa_content = {} - st_dic_master = {} - CTD_dic = {} - dp = {} - for i in range(len(self.df)): - i_name = self.df.index[i] - dna_feat[i_name] = count_orf(self.df.iloc[i]['DNAseq']) - aa_len[i_name] = len(self.df.iloc[i]['AAseq']) - aroma_dic[i_name] = ProteinAnalysis(self.df.iloc[i]['AAseq']).aromaticity() - iso_dic[i_name] = ProteinAnalysis(self.df.iloc[i]['AAseq']).isoelectric_point() - aa_content[i_name] = count_aa(self.df.iloc[i]['AAseq']) - st_dic_master[i_name] = sec_st_fr(self.df.iloc[i]['AAseq']) - CTD_dic[i_name] = CTD.CalculateCTD(self.df.iloc[i]['AAseq']) - dp[i_name] = AAComposition.CalculateDipeptideComposition(self.df.iloc[i]['AAseq']) - for j in self.df.index: - self.df.loc[j, dna_feat[j].keys()] = dna_feat[j].values() #dic with multiple values - self.df.loc[j, 'AA_Len'] = int(aa_len[j]) #dic with one value - self.df.loc[j, 'Aromaticity'] = aroma_dic[j] - self.df.loc[j, 'IsoelectricPoint'] = iso_dic[j] - self.df.loc[j, aa_content[j].keys()] = aa_content[j].values() - self.df.loc[j, st_dic_master[j].keys()] = st_dic_master[j].values() - self.df.loc[j, CTD_dic[j].keys()] = CTD_dic[j].values() - self.df.loc[j, dp[j].keys()] = dp[j].values() - self.df.drop(['DNAseq','AAseq'],axis=1,inplace=True) - - def Prediction(self): - import os - import pickle - import json - import pandas as pd - import numpy as np - from pathlib import Path - ft_scaler = pd.DataFrame(self.scaler.transform(self.df.iloc[:, :]), index=self.df.index,columns=self.df.columns) - ft_scaler = ft_scaler.drop(columns=[col for col in self.df if col not in self.feat[self.name]], axis=1) - scores = self.model.predict_proba(ft_scaler) - pos_scores = np.empty((self.df.shape[0], 0), float) - for x in scores: - pos_scores = np.append(pos_scores, round(x[1]*100)) - self.df_output.reset_index(inplace=True) - self.df_output['{} DPO Prediction (%)'.format(self.name)]= pos_scores - self.df_output = self.df_output.sort_values(by='{} DPO Prediction (%)'.format(self.name), ascending=False) - self.df_output.to_html('output.html', index=False, justify='center') - -if __name__ == '__main__': - import os - import sys - __location__ = os.path.realpath(os.path.join(os.getcwd(), os.path.dirname(__file__))) - - model = sys.argv[1] - fasta_file = sys.argv[2] - - PDPO = PDPOPrediction(__location__,model,fasta_file) - PDPO.Datastructure() - PDPO.Prediction() -
--- a/PhageDPO.xml Fri Nov 26 14:42:21 2021 +0000 +++ /dev/null Thu Jan 01 00:00:00 1970 +0000 @@ -1,68 +0,0 @@ -<tool id="PhageDPO" name="PhageDPO" version="0.1.0" python_template_version="3.5"> - <description> -Phage Depolymerase Finder - </description> - <requirements> - <requirement type="package" version="1.78">biopython</requirement> - <requirement type="package" version="0.24.1">scikit-learn</requirement> - <requirement type="package" version="1.21.2">numpy</requirement> - <requirement type="package" version="1.2.3">pandas</requirement> - <requirement type="package" version="1.0.2">propy3</requirement> - </requirements> - <command detect_errors="exit_code"><![CDATA[ - python '$__tool_directory__/DPOGALAXY.py' '$adv.model' '${input1}' - ]]></command> - <inputs> - <param type="data" name="input1" format="fasta" label="Fasta file"/> - <section name = 'adv' title= 'Advanced Options' expanded = 'False'> - <param type = "select" name="model" label="Model"> - <option value="SVM4311" selected="yes">SVM4311</option> - <option value="ANN4311">ANN4311</option> - <option value="RF5748">RF5748</option> - </param> - </section> - - </inputs> - <outputs> - <data name="output1" format="html" from_work_dir="output.html" - label="DPO Prediction"/> - </outputs> - <tests> - <test> - <param name="model" value="SVM4311"/> - <param name="input1" value="fasta_file.fasta"/> - <output name="output1" file="output.html"/> - </test> - </tests> - <help><![CDATA[ - -======== -PhageDPO -======== - -Predicts the existance of Phage Polysaccharide Depolymerase. - -PhageDPO is a python script that predicts the existance of depolymerases (DPOs) using supervised machine learning models. Two different datasets were used to develop three models: The SVM and ANN models was built using a dataset with 45 features and 4311 examples (1437 positives and 2874 negatives) and the RF model was created using a dataset with 54 features and 5748 examples (1437 positives and 4311 negatives). - -**Inputs:** - -* fasta file: fasta file format contain the ORF or the CDS. - -**Advanced options:** - -* Model: selection of the model to run: the SVM model (default), ANN model or the RF model. The SVM and ANN models focus on true positive detection. On the other hand, the RF model uses more negative data ensuring that all DPOs are identified, although with some misclassifications. - -**Outputs:** - -The tool outputs an html file containing the name of the sequence and the percentage of positive prediction for DPO. - -**Requirements:** - -* Biopython -* Sklearn -* Numpy -* Pandas -* Propy - - ]]></help> -</tool> \ No newline at end of file