Add MSEntropy score (https://github.com/YuanyueLi/MSEntropy)

libmetgem/__init__.py

@@ -4,27 +4,30 @@
 
 from .common import MZ, INTENSITY
 from ._cython import IS_CYTHONIZED
-from .filter import square_root_and_normalize_data
+from .filter import square_root_data, normalize_data
 
 import numpy as np
 
-def human_readable_data(data: np.ndarray) -> np.ndarray:
+def human_readable_data(data: np.ndarray, square_intensities=True) -> np.ndarray:
     """
         Normalize an MS/MS spectrum to the spectrum's maximum intensity and
         revert `filter_data` square-root process.
         
     Args:
         data: A 2D array representing an MS/MS spectrum.
+        square_intensities (bool): Whether to square intensities to revert
+            `filter_data` square-root process.
         
     Returns:
-        A copy of the array with intensities squared and normalised to maximum.
+        A copy of the array with intensities squared (optional) and normalised to maximum.
         
     See Also:
         filter_data
     """
 
     data = data.copy()
-    data[:, INTENSITY] = data[:, INTENSITY] ** 2
+    if square_intensities:
+        data[:, INTENSITY] = data[:, INTENSITY] ** 2
     data[:, INTENSITY] = data[:, INTENSITY] / data[:, INTENSITY].max() * 100
     return data
 

libmetgem/_common.pxd

@@ -15,7 +15,18 @@ ctypedef packed struct score_t:
     double value
     uint8_t type
 
+ctypedef enum score_algorithm_t:
+    cosine
+    entropy
+    weighted_entropy
+
+ctypedef enum norm_method_t:
+    dot
+    sum
+
 cdef np.ndarray[np.float32_t, ndim=2] arr_from_peaks_vector(vector[peak_t] v)
 cdef np.ndarray arr_from_score_vector(vector[score_t] v)
 cdef np.ndarray arr_from_1d_vector(vector[numeric] v, np.dtype dtype)
-cdef void *np_arr_pointer(np.ndarray[numeric, ndim=2] data)
+cdef void *np_arr_pointer(np.ndarray[numeric, ndim=2] data)
+cdef score_algorithm_t str_to_score_algorithm(str)
+cdef norm_method_t str_to_norm_method(str)

libmetgem/_common.pyx

@@ -92,3 +92,14 @@ cdef void *np_arr_pointer(np.ndarray[numeric, ndim=2] data):
     if not data.flags['C_CONTIGUOUS']:
         data = np.ascontiguousarray(data, dtype=data.dtype)
     return <void *>&data[0, 0]
+
+cdef score_algorithm_t str_to_score_algorithm(str score):
+    if score == 'weighted_entropy':
+        return score_algorithm_t.weighted_entropy
+    elif score == 'entropy':
+        return score_algorithm_t.entropy
+    else:
+        return score_algorithm_t.cosine
+
+cdef norm_method_t str_to_norm_method(str norm):
+    return norm_method_t.sum if norm == 'sum' else norm_method_t.dot

libmetgem/_cosine.pxd

@@ -2,12 +2,13 @@
 
 cimport numpy as np
 
-from ._common cimport peak_t
+from ._common cimport peak_t, score_algorithm_t
 
 cpdef enum SpectraMatchState:
     fragment = 0
     neutral_loss = 1
 
-cdef double cosine_score_nogil(double, peak_t*, np.npy_intp,
-                               double, peak_t*, np.npy_intp,
-                               double, int) noexcept nogil
+cdef double generic_score_nogil(
+    double, const peak_t*, np.npy_intp,
+    double, const peak_t*, np.npy_intp,
+    double, int, score_algorithm_t score_algorithm=*) noexcept nogil