From 93caebe123b084aa8b9056e9fe0f6ab356e7fe5e Mon Sep 17 00:00:00 2001
From: Fontrodona Nicolas <nicolas.fontrodona@ens-lyon.fr>
Date: Tue, 17 Nov 2020 15:11:55 +0100
Subject: [PATCH] src/find_interaction_cluster/radomization_test_ppi.py: add 3
functions to check if the number communities that are enriched in common
genes between dna and protein level is also enriched
---
.../radomization_test_ppi.py | 233 +++++++++++++++++-
1 file changed, 220 insertions(+), 13 deletions(-)
diff --git a/src/find_interaction_cluster/radomization_test_ppi.py b/src/find_interaction_cluster/radomization_test_ppi.py
index fa046d4a..4a7e2e95 100644
--- a/src/find_interaction_cluster/radomization_test_ppi.py
+++ b/src/find_interaction_cluster/radomization_test_ppi.py
@@ -90,8 +90,8 @@ def get_common_genes(dna_gene: np.array, ppi_gene: np.array) -> int:
def get_list_common_gene(community: str, dic_dna_gene: Dict[str, np.array],
- ppi_size: int, ppi_gene: np.array, iteration: int
- ) -> np.array:
+ ppi_size: int, ppi_gene: np.array, iteration: int,
+ use_seed: bool = True) -> np.array:
"""
Return the number of common gene between genes in the community \
`community`and a sublist of randomly chosen genes in `ppi_gene` of \
@@ -103,6 +103,7 @@ def get_list_common_gene(community: str, dic_dna_gene: Dict[str, np.array],
:param ppi_size: The size of the ppi_gene subsample
:param ppi_gene: All the gene interacting at protein level
:param iteration: The number of time we want to repeat the process
+ :param use_seed: True to use a fixed seed, false else.
:return: The number of gene in common between the genes in community \
`community` and the subsample of ppi_size
@@ -119,7 +120,8 @@ def get_list_common_gene(community: str, dic_dna_gene: Dict[str, np.array],
>>> get_list_common_gene("C1", dna_gene, 10, ppi_g, 5)
array([5, 5, 5, 5, 5])
"""
- np.random.seed(1)
+ if use_seed:
+ np.random.seed(1)
common = [
get_common_genes(
dic_dna_gene[community],
@@ -180,11 +182,24 @@ def get_pvalue(values: np.array, target: float, iteration: int
return pval, regulation
+def adjust_regulation(x: pd.Series) -> str:
+ """
+ Return the adjusted regulation.
+
+ :param x: A row a the dataframe
+ :return: The adjusted row
+ """
+ return x["regulation"] if x["p-adjust"] <= 0.05 else " . "
+
+
def update_overlap_df(df_overlap: pd.DataFrame,
dic_dna_gene: Dict[str, np.array],
- ppi_gene: np.array, iteration: int
- ) -> pd.DataFrame:
+ ppi_gene: np.array, iteration: int,
+ use_seed: bool = True,
+ ) -> Tuple[pd.DataFrame, np.array]:
"""
+ Perform a randomization test to check if communities at \
+ DNA level are more similar at protein level than randomly expected.
:param df_overlap: A dataframe containing every gene community at dna \
level and the number of common gene with a gne community at protein \
@@ -193,36 +208,48 @@ def update_overlap_df(df_overlap: pd.DataFrame,
level
:param ppi_gene: All the gene interacting at protein level
:param iteration: The number of time we want to repeat the process
- :return: The updated Df (column p-value, p-adjust and regulation
+ :param use_seed: True to use a fixed seed, false else.
+ :return: The updated Df (column p-value, p-adjust and regulation) + \
+ the list of common gene shared between gene community at DNA and \
+ protein level obtained by randmization analysis
>>> do = pd.DataFrame({"DNA_community": ["C1", "C2", "C3"],
... "community_size": [2, 3, 2], "nb_com-ppi": [2, 3, 1],
... "size_com-ppi": [10, 10, 10]})
>>> dna_gene = {"C1": [1, 2], "C2": [3, 4, 5], "C3": [6, 7]}
>>> ppi_g = list(range(11))
- >>> update_overlap_df(do, dna_gene, ppi_g, 100).iloc[:,
- ... [0, 1, 2, 4, 5]]
+ >>> d, di = update_overlap_df(do, dna_gene, ppi_g, 100)
+ >>> d.iloc[:, [0, 1, 2, 4, 5]]
DNA_community community_size nb_com-ppi p-value p-adjust
0 C1 2 2 0.79 0.79
1 C2 3 3 0.72 0.79
2 C3 2 1 0.18 0.54
- >>> d = update_overlap_df(do, dna_gene, list(range(1000)), 100)
+ >>> d.iloc[:, 6].to_list() == d.iloc[:, 7].to_list() == [" . "] * 3
+ True
+ >>> d, di = update_overlap_df(do, dna_gene, list(range(1000)), 100)
>>> d.iloc[:, [0, 1, 2, 4, 5]]
DNA_community community_size nb_com-ppi p-value p-adjust
0 C1 2 2 0.01 0.01
1 C2 3 3 0.00 0.00
2 C3 2 1 0.00 0.00
- >>> d["regulation"].to_list() == [" + "] * 3
+ >>> d.iloc[:, 6].to_list() == d.iloc[:, 7].to_list() == [" + "] * 3
+ True
+ >>> len(di.keys()) == 3
+ True
+ >>> len(di[list(di.keys())[0]]) == 100
True
"""
pval_list = []
reg_list = []
+ dic = {}
for i in tqdm(range(df_overlap.shape[0])):
- values = get_list_common_gene(df_overlap.iloc[i, :]["DNA_community"],
+ community = df_overlap.iloc[i, :]["DNA_community"]
+ values = get_list_common_gene(community,
dic_dna_gene,
df_overlap.iloc[i, :]["size_com-ppi"],
ppi_gene,
- iteration)
+ iteration, use_seed)
+ dic[community] = values
pval, reg = get_pvalue(values, df_overlap.iloc[i, :]["nb_com-ppi"],
iteration)
pval_list.append(pval)
@@ -233,7 +260,187 @@ def update_overlap_df(df_overlap: pd.DataFrame,
is_sorted=False,
returnsorted=False)[1]
df_overlap["regulation"] = reg_list
- return df_overlap
+ df_overlap["regulation-adjusted"] = \
+ df_overlap.apply(adjust_regulation, axis=1)
+ return df_overlap, dic
+
+
+def summarize_df_overlap(df_overlap: pd.DataFrame) -> pd.DataFrame:
+ """
+ Create a summary of the dataframe `df_overlap` that indicates the number \
+ of common gene shared between communities at protein and DNA level.
+
+ :param df_overlap: The dataframe containing community at DNA and \
+ protein level and their enrichment
+ :return: The dataframe showing the number of enrichment, impoverishment \
+ or no regulation
+
+ >>> do = pd.DataFrame({"DNA_community": ["C1", "C2", "C3"],
+ ... "community_size": [2, 3, 2], "nb_com-ppi": [2, 3, 1],
+ ... "size_com-ppi": [10, 10, 10], "p-value": [0.01, 0, 0],
+ ... "p-adjust": [0.01, 0.01, 0.01], "regulation": [" + "] * 3,
+ ... "regulation-adjusted": [" + "] * 3})
+ >>> summarize_df_overlap(do)
+ regulation-adjusted community_size nb_com-ppi size_com-ppi number
+ 0 + 2.333333 2.0 10.0 3
+ 1 - NaN NaN NaN 0
+ 2 . NaN NaN NaN 0
+ """
+ df_overlap = df_overlap[["regulation-adjusted", "community_size",
+ "nb_com-ppi", "size_com-ppi", "p-adjust"]]
+ df_tmp = pd.DataFrame({"regulation-adjusted": [" + ", " - ", " . "],
+ "community_size": [np.nan] * 3,
+ "nb_com-ppi": [np.nan] * 3,
+ "size_com-ppi": [np.nan] * 3,
+ "p-adjust": [np.nan] * 3})
+ df = pd.concat([df_overlap, df_tmp], axis=0, ignore_index=True)
+ df = df.groupby("regulation-adjusted").agg({"community_size": "mean",
+ "nb_com-ppi": "mean",
+ "size_com-ppi": "mean",
+ "p-adjust": "count"})
+ return df.rename({"p-adjust": "number"}, axis=1).reset_index()
+
+
+def get_enriched_impoverished_communities(df_overlap: pd.DataFrame,
+ dic_dna_gene: Dict[str, np.array],
+ ppi_gene: np.array, iteration: int,
+ dic_values: Dict[str, np.array],
+ use_seed: bool = True
+ ) -> Dict[str, int]:
+ """
+ For each communities, randomly samples only once in the list \
+ of `ppi_gene` to get the number of common gene betwwen this sample \
+ and the community of gene at DNA level. Then this function determines \
+ how many communites are enriched, impovershied thanks to `di_values`.
+
+ :param df_overlap: A dataframe containing every gene community at dna \
+ level and the number of common gene with a gne community at protein \
+ level
+ :param dic_dna_gene: The dictionary containing gene interacting at dna \
+ level
+ :param ppi_gene: All the gene interacting at protein level
+ :param iteration: The number of time we want to repeat the process
+ :param dic_values: A dictionary containing the number of common \
+ gene between communities at DNA and protein level for each \
+ subsampling made from the ppi_gene list.
+ :param use_seed: True to use a fixed seed, false else.
+ :return: The number of enriched, impoverished number of \
+ genes shared betwwen gene at DNA and protein level.
+
+ >>> do = pd.DataFrame({"DNA_community": ["C1", "C2", "C3"],
+ ... "community_size": [2, 3, 2], "nb_com-ppi": [2, 3, 1],
+ ... "size_com-ppi": [10, 10, 10], "p-value": [0.01, 0, 0],
+ ... "p-adjust": [0.01, 0.01, 0.01], "regulation": [" + "] * 3,
+ ... "regulation-adjusted": [" + "] * 3})
+ >>> dna_gene = {"C1": [1, 2], "C2": [3, 4, 5], "C3": [6, 7]}
+ >>> ppi_g = list(range(11))
+ >>> dv = {"C1": [0] * 95 + [1, 1, 1, 2, 2],
+ ... "C2": [0] * 95 + [1, 1, 1, 2, 2],
+ ... "C3": [0] * 75 + [1] * 10 + [2] * 10 + [3] * 5}
+ >>> get_enriched_impoverished_communities(do, dna_gene, ppi_g, 100, dv)
+ {' + ': 2, ' . ': 1, ' - ': 0}
+ >>> dv = {"C1": [0] * 95 + [1, 1, 1, 2, 2],
+ ... "C2": [0] * 95 + [1, 1, 1, 2, 2],
+ ... "C3": [3] * 75 + [4] * 23 + [1] * 2}
+ >>> get_enriched_impoverished_communities(do, dna_gene, ppi_g, 100, dv)
+ {' + ': 2, ' . ': 0, ' - ': 1}
+ """
+ pval_list = []
+ reg_list = []
+ for i in range(df_overlap.shape[0]):
+ community = df_overlap.iloc[i, :]["DNA_community"]
+ my_value = get_list_common_gene(community,
+ dic_dna_gene,
+ df_overlap.iloc[i, :]["size_com-ppi"],
+ ppi_gene,
+ 1, use_seed)[0]
+ pval, reg = get_pvalue(dic_values[community], my_value,
+ iteration)
+ pval_list.append(pval)
+ reg_list.append(reg)
+ p_adjust = multipletests(pval_list, alpha=0.05, method='fdr_bh',
+ is_sorted=False, returnsorted=False)[1]
+ dic = {" + ": 0, " . ": 0, " - ": 0}
+ for i, reg in enumerate(reg_list):
+ dic[reg if p_adjust[i] <= 0.05 else " . "] += 1
+ return dic
+
+
+def summary_randomisation_test(df_overlap: pd.DataFrame,
+ dic_dna_gene: Dict[str, np.array],
+ ppi_gene: np.array, iteration: int,
+ dic_values: Dict[str, np.array],
+ use_seed: bool = True
+ ) -> pd.DataFrame:
+ """
+ Says if the number of community that shared an enriched number \
+ of common gene between the protein and DNA level is enriched as well.
+
+ :param df_overlap: A dataframe containing every gene community at dna \
+ level and the number of common gene with a gne community at protein \
+ level
+ :param dic_dna_gene: The dictionary containing gene interacting at dna \
+ level
+ :param ppi_gene: All the gene interacting at protein level
+ :param iteration: The number of time we want to repeat the process
+ :param dic_values: A dictionary containing the number of common \
+ gene between communities at DNA and protein level for each \
+ subsampling made from the ppi_gene list.
+ :param use_seed: True to use a fixed seed, false else.
+ :return: The updated summary table
+
+ >>> do = pd.DataFrame({"DNA_community": ["C1", "C2", "C3"],
+ ... "community_size": [2, 3, 2], "nb_com-ppi": [2, 3, 1],
+ ... "size_com-ppi": [10, 10, 10], "p-value": [0.01, 0, 0],
+ ... "p-adjust": [0.01, 0.01, 0.01], "regulation": [" + "] * 3,
+ ... "regulation-adjusted": [" + "] * 3})
+ >>> dna_gene = {"C1": [1, 2], "C2": [3, 4, 5], "C3": [6, 7]}
+ >>> ppi_g = list(range(11))
+ >>> dv = {"C1": [0] * 95 + [1, 1, 1, 2, 2],
+ ... "C2": [0] * 95 + [1, 1, 1, 2, 2],
+ ... "C3": [0] * 75 + [1] * 10 + [2] * 10 + [3] * 5}
+ >>> ds = summary_randomisation_test(do, dna_gene, ppi_g, 100, dv)
+ >>> ds.iloc[:, range(4)]
+ regulation-adjusted community_size nb_com-ppi size_com-ppi
+ 0 + 2.333333 2.0 10.0
+ 1 - NaN NaN NaN
+ 2 . NaN NaN NaN
+ >>> ds.iloc[:, range(4, 7)]
+ number mean_100_number p-adjust
+ 0 3 2.0 0.0
+ 1 0 0.0 1.0
+ 2 0 1.0 0.0
+ >>> ds.iloc[:, 7].to_list() == [" + ", " . ", " - "]
+ True
+ """
+ df_summarized = summarize_df_overlap(df_overlap)
+ dic_rand_sum = {k: [] for k in df_summarized["regulation-adjusted"].values}
+ for _ in tqdm(range(iteration), desc="Making stats to summarise global "
+ "enrichment"):
+ d = get_enriched_impoverished_communities(df_overlap, dic_dna_gene,
+ ppi_gene, iteration,
+ dic_values, use_seed)
+ for k, v in dic_rand_sum.items():
+ v.append(d[k])
+ pvalues = []
+ regulations = []
+ mean_number = []
+ for i in range(df_summarized.shape[0]):
+ reg, val = df_summarized.iloc[i, :][["regulation-adjusted", "number"]]
+ pval, creg = get_pvalue(dic_rand_sum[reg], val,
+ iteration)
+ # print(reg, val, dic_rand_sum[reg])
+ pvalues.append(pval)
+ regulations.append(creg)
+ mean_number.append(np.mean(dic_rand_sum[reg]))
+ p_adjust = multipletests(pvalues, alpha=0.05, method='fdr_bh',
+ is_sorted=False, returnsorted=False)[1]
+ reg_adj = [regulations[i] if p_adjust[i] <= 0.05 else " . "
+ for i in range(len(regulations))]
+ df_summarized[f"mean_{iteration}_number"] = mean_number
+ df_summarized[f"p-adjust"] = p_adjust
+ df_summarized[f"reg_adj"] = reg_adj
+ return df_summarized
if __name__ == "__main__":
--
GitLab