option to treat y_s as detetion probability

Author: vitkl

cell2location/models/_cell2location_model.py

@@ -76,11 +76,15 @@ def __init__(
         self.factor_names_ = cell_state_df.columns.values
 
         if not detection_mean_per_sample:
-            # compute expected change in sensitivity (m_g in V1 or y_s in V2)
+            # compute expected change in sensitivity (m_g in V1 and y_s in V2)
             sc_total = cell_state_df.sum(0).mean()
-            sp_total = get_from_registry(self.adata, _CONSTANTS.X_KEY).sum(1).mean()
-            get_from_registry(adata, _CONSTANTS.BATCH_KEY)
+            sp_total = get_from_registry(self.adata, _CONSTANTS.X_KEY).sum(1)
+            batch = get_from_registry(self.adata, _CONSTANTS.BATCH_KEY).flatten()
+            sp_total = np.array([sp_total[batch == b].mean() for b in range(self.summary_stats["n_batch"])])
             self.detection_mean_ = (sp_total / model_kwargs.get("N_cells_per_location", 1)) / sc_total
+            if (self.detection_mean_.max() > 1.0) and (model_kwargs.get("use_detection_probability", False) is True):
+                self.detection_mean_ = self.detection_mean_ / (self.detection_mean_.max() + 0.000001)
+            self.detection_mean_ = self.detection_mean_.mean()
             self.detection_mean_ = self.detection_mean_ * detection_mean_correction
             model_kwargs["detection_mean"] = self.detection_mean_
         else:

cell2location/models/_cell2location_module.py

@@ -83,6 +83,7 @@ def __init__(
         n_groups: int = 50,
         detection_mean=1 / 2,
         detection_alpha=200.0,
+        use_detection_probability: bool = False,
         m_g_gene_level_prior={"mean": 1, "mean_var_ratio": 1.0, "alpha_mean": 3.0},
         N_cells_per_location=8.0,
         A_factors_per_location=7.0,
@@ -117,6 +118,7 @@ def __init__(
         detection_hyp_prior["mean"] = detection_mean
         detection_hyp_prior["alpha"] = detection_alpha
         self.detection_hyp_prior = detection_hyp_prior
+        self.use_detection_probability = use_detection_probability
 
         if (init_vals is not None) & (type(init_vals) is dict):
             self.np_init_vals = init_vals
@@ -323,27 +325,47 @@ def forward(self, x_data, idx, batch_index):
                 )  # (self.n_obs, self.n_factors)
 
         # =====================Location-specific detection efficiency ======================= #
-        # y_s with hierarchical mean prior
-        detection_mean_y_e = pyro.sample(
-            "detection_mean_y_e",
-            dist.Gamma(
-                self.ones * self.detection_mean_hyp_prior_alpha,
-                self.ones * self.detection_mean_hyp_prior_beta,
+        if not self.use_detection_probability:
+            # y_s with hierarchical mean prior
+            detection_mean_y_e = pyro.sample(
+                "detection_mean_y_e",
+                dist.Gamma(
+                    self.ones * self.detection_mean_hyp_prior_alpha,
+                    self.ones * self.detection_mean_hyp_prior_beta,
+                )
+                .expand([self.n_batch, 1])
+                .to_event(2),
+            )
+            detection_hyp_prior_alpha = pyro.deterministic(
+                "detection_hyp_prior_alpha",
+                self.ones_n_batch_1 * self.detection_hyp_prior_alpha,
             )
-            .expand([self.n_batch, 1])
-            .to_event(2),
-        )
-        detection_hyp_prior_alpha = pyro.deterministic(
-            "detection_hyp_prior_alpha",
-            self.ones_n_batch_1 * self.detection_hyp_prior_alpha,
-        )
 
-        beta = (obs2sample @ detection_hyp_prior_alpha) / (obs2sample @ detection_mean_y_e)
-        with obs_plate:
-            detection_y_s = pyro.sample(
-                "detection_y_s",
-                dist.Gamma(obs2sample @ detection_hyp_prior_alpha, beta),
-            )  # (self.n_obs, 1)
+            beta = (obs2sample @ detection_hyp_prior_alpha) / (obs2sample @ detection_mean_y_e)
+            with obs_plate:
+                detection_y_s = pyro.sample(
+                    "detection_y_s",
+                    dist.Gamma(obs2sample @ detection_hyp_prior_alpha, beta),
+                )  # (self.n_obs, 1)
+        else:
+            # y_s with hierarchical mean prior
+            detection_mean_y_e = pyro.sample(
+                "detection_mean_y_e",
+                dist.Beta(
+                    self.ones * self.detection_mean_hyp_prior_alpha,
+                    self.ones * self.detection_mean_hyp_prior_beta,
+                )
+                .expand([self.n_batch, 1])
+                .to_event(2),
+            )
+
+            alpha = (obs2sample @ detection_mean_y_e) * self.ones * self.detection_hyp_prior_alpha
+            beta = (obs2sample @ (self.ones - detection_mean_y_e)) * self.ones * self.detection_hyp_prior_alpha
+            with obs_plate:
+                detection_y_s = pyro.sample(
+                    "detection_y_s",
+                    dist.Beta(alpha, beta),
+                )  # (self.n_obs, 1)
 
         # =====================Gene-specific additive component ======================= #
         # per gene molecule contribution that cannot be explained by

tests/test_cell2location.py

@@ -120,3 +120,19 @@ def test_cell2location():
     # export the estimated cell abundance (summary of the posterior distribution)
     # full data
     st_model.export_posterior(dataset, sample_kwargs={"num_samples": 10, "batch_size": st_model.adata.n_obs})
+
+    ### test new cell2location models ###
+    ##  detection probability rather than detection efficiency  ##
+    st_model = Cell2location(
+        dataset,
+        cell_state_df=inf_aver,
+        N_cells_per_location=30,
+        detection_alpha=200,
+        use_detection_probability=True,
+        detection_hyp_prior={"mean_alpha": 100.0},
+    )
+    # test full data training
+    st_model.train(max_epochs=1)
+    # export the estimated cell abundance (summary of the posterior distribution)
+    # full data
+    dataset = st_model.export_posterior(dataset, sample_kwargs={"num_samples": 10, "batch_size": st_model.adata.n_obs})