experiment.py 文件源码

def calc_pvalues_wt(self, label):
        """
        Calculate uncorrected pvalue for each variant compared to wild type.
        """
        if self.check_store("/main/{}/scores_pvalues_wt".format(label)):
            return

        idx = pd.IndexSlice

        wt = self.store.select("/main/{}/scores".format(label),
                               "index=WILD_TYPE_VARIANT")
        if len(wt) == 0:    # no wild type score
            logging.info("Failed to find wild type score, skipping wild type "
                         "p-value calculations", extra={'oname': self.name})
            return
        data = self.store.select("/main/{}/scores".format(label),
                                 "index!=WILD_TYPE_VARIANT")

        columns = pd.MultiIndex.from_product([sorted(self.child_names()),
                                              sorted(["z", "pvalue_raw"])],
                                             names=["condition", "value"])
        result_df = pd.DataFrame(index=data.index, columns=columns)

        condition_labels = data.columns.levels[0]
        for cnd in condition_labels:
            result_df.loc[:, idx[cnd, 'z']] = \
                np.absolute(wt.loc[WILD_TYPE_VARIANT, idx[cnd, 'score']] -
                            data.loc[:, idx[cnd, 'score']]) / \
                np.sqrt(wt.loc[WILD_TYPE_VARIANT, idx[cnd, 'SE']] ** 2 +
                        data.loc[:, idx[cnd, 'SE']] ** 2)
            result_df.loc[:, idx[cnd, 'pvalue_raw']] = \
                2 * stats.norm.sf(result_df.loc[:, idx[cnd, 'z']])

        self.store.put("/main/{}/scores_pvalues_wt".format(label), result_df,
                       format="table")