Fix running without calibration

Author: rodvrees

im2deep/__init__.py

@@ -1,3 +1,3 @@
 """IM2Deep: Deep learning framework for peptide collisional cross section prediction."""
 
-__version__ = "0.1.5"
+__version__ = "0.1.6"

im2deep/__main__.py

@@ -104,18 +104,17 @@ def setup_logging(passed_level):
 @click.option(
     "--use_single_model",
     type=click.BOOL,
-    default=False,
+    default=True,
     help="Use a single model for prediction.",
 )
-
 def main(
     psm_file: str,
     calibration_file: Optional[str] = None,
     output_file: Optional[str] = None,
     model_name: Optional[str] = "tims",
     log_level: Optional[str] = "info",
     n_jobs: Optional[int] = None,
-    use_single_model: Optional[bool] = False,
+    use_single_model: Optional[bool] = True,
     calibrate_per_charge: Optional[bool] = True,
     use_charge_state: Optional[int] = 2,
 ):
@@ -156,31 +155,38 @@ def main(
         and "modifications" in first_line_cal.split(",")
         and "seq" in first_line_cal.split(",")
     ):
-        df_cal = pd.read_csv(calibration_file)
-        df_cal.fillna("", inplace=True)
-        del calibration_file
-
-        list_of_cal_psms = []
-        for seq, mod, charge, ident, CCS in zip(
-            df_cal["seq"],
-            df_cal["modifications"],
-            df_cal["charge"],
-            df_cal.index,
-            df_cal["CCS"],
-        ):
-            list_of_cal_psms.append(
-                PSM(peptidoform=peprec_to_proforma(seq, mod, charge), spectrum_id=ident)
+        try:
+            df_cal = pd.read_csv(calibration_file)
+            df_cal.fillna("", inplace=True)
+            del calibration_file
+
+            list_of_cal_psms = []
+            for seq, mod, charge, ident, CCS in zip(
+                df_cal["seq"],
+                df_cal["modifications"],
+                df_cal["charge"],
+                df_cal.index,
+                df_cal["CCS"],
+            ):
+                list_of_cal_psms.append(
+                    PSM(peptidoform=peprec_to_proforma(seq, mod, charge), spectrum_id=ident)
+                )
+            psm_list_cal = PSMList(psm_list=list_of_cal_psms)
+            psm_list_cal_df = psm_list_cal.to_dataframe()
+            psm_list_cal_df["ccs_observed"] = df_cal["CCS"]
+            del df_cal
+
+        except IOError:
+            LOGGER.error(
+                "Invalid calibration file. Please check the format of the calibration file."
             )
-        psm_list_cal = PSMList(psm_list=list_of_cal_psms)
-        psm_list_cal_df = psm_list_cal.to_dataframe()
-        psm_list_cal_df["ccs_observed"] = df_cal["CCS"]
-        del df_cal
+            sys.exit(1)
 
     else:
-        LOGGER.error(
-            "Invalid calibration file. Please check the format of the calibration file."
+        LOGGER.warning(
+            "No calibration file found. Proceeding without calibration. Calibration is HIGHLY recommended for accurate CCS prediction."
         )
-        sys.exit(1)
+        psm_list_cal_df = None
 
     if not output_file:
         output_file = Path(psm_file).parent / (Path(psm_file).stem + "_IM2Deep-predictions.csv")

im2deep/calibrate.py

@@ -7,6 +7,7 @@
 
 LOGGER = logging.getLogger(__name__)
 
+
 def im2ccs(reverse_im, mz, charge, mass_gas=28.013, temp=31.85, t_diff=273.15):
     """
     Convert ion mobility to collisional cross section.
@@ -82,7 +83,9 @@ def get_ccs_shift(
     )
     LOGGER.debug(
         """Calculating CCS shift based on {} overlapping peptide-charge pairs
-        between PSMs and reference dataset""".format(both.shape[0])
+        between PSMs and reference dataset""".format(
+            both.shape[0]
+        )
     )
 
     # How much CCS in calibration data is larger than reference CCS, so predictions
@@ -156,7 +159,7 @@ def calculate_ccs_shift(
         CCS shift factor.
 
     """
-    cal_df['charge'] = cal_df['peptidoform'].apply(lambda x: x.precursor_charge)
+    cal_df["charge"] = cal_df["peptidoform"].apply(lambda x: x.precursor_charge)
     cal_df = cal_df[cal_df["charge"] < 7]  # predictions do not go higher for IM2Deep
 
     if not per_charge:
@@ -204,27 +207,36 @@ def linear_calibration(
 
     """
     LOGGER.info("Calibrating CCS values using linear calibration...")
-    preds_df["charge"] = preds_df["peptidoform"].apply(lambda x: x.precursor_charge)
     if per_charge:
         general_shift = calculate_ccs_shift(
-            calibration_dataset, reference_dataset, per_charge=False, use_charge_state=use_charge_state
+            calibration_dataset,
+            reference_dataset,
+            per_charge=False,
+            use_charge_state=use_charge_state,
         )
         shift_factor_dict = calculate_ccs_shift(
             calibration_dataset, reference_dataset, per_charge=True
         )
         for charge in preds_df["charge"].unique():
             if charge not in shift_factor_dict:
-                LOGGER.info("No overlapping precursors for charge state {}. Using overall shift factor for precursors with that charge.".format(charge))
+                LOGGER.info(
+                    "No overlapping precursors for charge state {}. Using overall shift factor for precursors with that charge.".format(
+                        charge
+                    )
+                )
                 shift_factor_dict[charge] = general_shift
         LOGGER.info("Shift factors per charge: {}".format(shift_factor_dict))
-        preds_df["predicted_ccs_calibrated"] = preds_df.apply(
+        preds_df["predicted_ccs"] = preds_df.apply(
             lambda x: x["predicted_ccs"] + shift_factor_dict[x["charge"]], axis=1
         )
     else:
         shift_factor = calculate_ccs_shift(
-            calibration_dataset, reference_dataset, per_charge=False, use_charge_state=use_charge_state
+            calibration_dataset,
+            reference_dataset,
+            per_charge=False,
+            use_charge_state=use_charge_state,
         )
-        preds_df["predicted_ccs_calibrated"] = preds_df.apply(
+        preds_df["predicted_ccs"] = preds_df.apply(
             lambda x: x["predicted_ccs"] + shift_factor, axis=1
         )
     LOGGER.info("CCS values calibrated.")

im2deep/im2deep.py

@@ -20,7 +20,7 @@ def predict_ccs(
     model_name="tims",
     calibrate_per_charge=True,
     use_charge_state=2,
-    use_single_model=False,
+    use_single_model=True,
     n_jobs=None,
     write_output=True,
 ):
@@ -33,33 +33,40 @@ def predict_ccs(
 
     path_model_list = list(path_model.glob("*.hdf5"))
     if use_single_model:
-        path_model_list = [path_model_list[0]]
+        path_model_list = [path_model_list[1]]
 
     dlc = DeepLC(path_model=path_model_list, n_jobs=n_jobs, predict_ccs=True)
     LOGGER.info("Predicting CCS values...")
     preds = dlc.make_preds(psm_list=psm_list_pred, calibrate=False)
     LOGGER.info("CCS values predicted.")
     psm_list_pred_df = psm_list_pred.to_dataframe()
     psm_list_pred_df["predicted_ccs"] = preds
-
-    calibrated_psm_list_pred_df = linear_calibration(
-        psm_list_pred_df,
-        calibration_dataset=psm_list_cal_df,
-        reference_dataset=reference_dataset,
-        per_charge=calibrate_per_charge,
-        use_charge_state=use_charge_state,
+    psm_list_pred_df["charge"] = psm_list_pred_df["peptidoform"].apply(
+        lambda x: x.precursor_charge
     )
+
+    if psm_list_cal_df is not None:
+        psm_list_pred_df = linear_calibration(
+            psm_list_pred_df,
+            calibration_dataset=psm_list_cal_df,
+            reference_dataset=reference_dataset,
+            per_charge=calibrate_per_charge,
+            use_charge_state=use_charge_state,
+        )
+
+    LOGGER.debug(psm_list_pred_df)
     if write_output:
         LOGGER.info("Writing output file...")
         output_file = open(output_file, "w")
         output_file.write("seq,modifications,charge,predicted CCS\n")
         for peptidoform, charge, CCS in zip(
-            calibrated_psm_list_pred_df["peptidoform"],
-            calibrated_psm_list_pred_df["charge"],
-            calibrated_psm_list_pred_df["predicted_ccs_calibrated"],
+            psm_list_pred_df["peptidoform"],
+            psm_list_pred_df["charge"],
+            psm_list_pred_df["predicted_ccs"],
         ):
             output_file.write(f"{peptidoform},{charge},{CCS}\n")
         output_file.close()
 
     LOGGER.info("IM2Deep finished!")
-    return calibrated_psm_list_pred_df["predicted_ccs_calibrated"]
+
+    return psm_list_pred_df["predicted_ccs"]