#!/usr/bin/env python
# coding: utf-8

# # New Mapping File from NCBI Gene Dataset
# Downloaded: Jan 2, 2019

# In[1]:


import csv
import pandas as pd


# In[2]:


file =  "/Users/maayan/sigsets/Harmonizome/Data/All_Data.gene_info_010219"
mapping_file  = "/Users/maayan/sigsets/Harmonizome/Data/mappingFileHMR_2019.tsv"
gene_sym_ids = "/Users/maayan/sigsets/Harmonizome/Data/GeneSymbolsAndIDSHMR_2019.tsv"


# In[3]:


# Human, Mouse, and Rat only
mapping_dict = {}
with open(file) as o:
    with open(gene_sym_ids, "w") as w:
        with open(mapping_file, "w") as m:
            csv_file = csv.reader(o, delimiter="\t")
            w.write("Human, Mouse, and Rat Approved Symbol	Entrez Gene ID(supplied by NCBI)")
            for row in csv_file:
                tax_id = row[0]
                if tax_id in ["9606", "10090", "10116"]:
                    gene_id = row[1]
                    if not row[2] == "NEWENTRY":
                        sym = row[2]
                    else: 
                        continue
                    synonyms = [sym]
                    if not row[4] == "-":
                        synonyms = synonyms + row[4].split("|")
                else:
                    continue
                w.write("\t".join([tax_id, sym, gene_id]) + "\n") # Add Taxon ID
                if not tax_id in mapping_dict:
                    mapping_dict[tax_id] = set([])
                for syn in synonyms:
                    if syn not in mapping_dict[tax_id]:
                        mapping_dict[tax_id].add(syn)
                        m.write("\t".join([tax_id, syn, sym])+ "\n")
                        


# In[4]:


getGeneIDsHMR_updated = pd.read_csv('/Users/maayan/sigsets/Harmonizome/Data/GeneSymbolsAndIDSHMR_2019.tsv', sep='\t', index_col=[0,1]).sort_index()


# In[14]:


(9606,"A2MP1") in getGeneIDsHMR_updated.index


# In[17]:


mappingDFHMR_updated = pd.read_csv('/Users/maayan/sigsets/Harmonizome/Data/mappingFileHMR_2019.tsv', sep='\t', header=None, index_col=[0,1]).sort_index()


# In[18]:


(9606,"(FM-3)") in mappingDFHMR_updated.index


# In[8]:


mappingDFHMR_updated


# In[ ]: