Skip to content
Merged
Show file tree
Hide file tree
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
2 changes: 2 additions & 0 deletions src/mzml_utils/__init__.py
Original file line number Diff line number Diff line change
Expand Up @@ -104,6 +104,7 @@
PeakMetrics,
find_peaks,
TracePeak,
bridge_dropouts,
beta_values,
)

Expand Down Expand Up @@ -226,6 +227,7 @@
"PeakMetrics",
"find_peaks",
"TracePeak",
"bridge_dropouts",
"beta_values",
# chrom_extract (peak table -> chromatograms across runs)
"extract_chromatograms",
Expand Down
46 changes: 38 additions & 8 deletions src/mzml_utils/xic.py
Original file line number Diff line number Diff line change
Expand Up @@ -459,6 +459,30 @@ def _bounds_from_apex(inten: np.ndarray, apex: int, baseline: float, threshold:
return int(left), int(right)


def bridge_dropouts(intensity, max_gap: int = 1) -> np.ndarray:
"""A copy of ``intensity`` with every run of at most ``max_gap`` zero scans that has signal
on both sides filled by linear interpolation between its neighbours.

A zero-filled XIC writes 0 for a scan without the ion. Near the Orbitrap's noise floor a
single scan inside an elution often lacks the centroid, and every valley walk stops there,
splitting one elution in two. Bridging rejoins it. Interpolated values lie between their
neighbours, so no maximum is created or moved. Leading and trailing zeros, and longer gaps,
are left as they are. ``max_gap=0`` returns an unchanged copy.
"""
return _bridge_gaps(np.asarray(intensity, dtype=float), int(max_gap))


def _bridge_gaps(raw: np.ndarray, max_gap: int) -> np.ndarray:
out = raw.copy()
if max_gap <= 0:
return out
nz = np.flatnonzero(raw > 0)
for a, b in zip(nz[:-1], nz[1:]):
if 1 < b - a <= max_gap + 1:
out[a + 1:b] = np.interp(np.arange(a + 1, b), [a, b], [raw[a], raw[b]])
return out


def peak_boundary(xic: "XIC", *,
contains_rt: Optional[float] = None,
apex_window: Optional[Tuple[float, float]] = None,
Expand Down Expand Up @@ -670,7 +694,8 @@ def peak_metrics(xic: "XIC", *,
threshold: float = 0.1,
baseline_threshold: float = 0.1,
baseline_quantile: float = 0.1,
nearest_valley: bool = False) -> PeakMetrics:
nearest_valley: bool = False,
bridge_dropouts: int = 0) -> PeakMetrics:
"""Boundaries, in-boundary area and quality metrics of one peak in an XIC.

The boundary arguments are those of :func:`peak_boundary`: ``contains_rt`` (or
Expand All @@ -687,8 +712,18 @@ def peak_metrics(xic: "XIC", *,
half (MAD) or a tenth (q10) of the trace is zero, as MsQuality returns NA --
common for a zero-filled XIC over a wide window.

``bridge_dropouts=k`` measures the trace after :func:`bridge_dropouts` (runs of at most k
zero scans with signal on both sides filled), as :func:`find_peaks` does, so the two agree
on one zero-filled trace; every metric then comes from the bridged trace except
``n_signal``, which counts the scans where the ORIGINAL trace has the ion. The default 0
reproduces R.

These are measurements, not verdicts: no threshold is applied here.
"""
raw = np.asarray(xic.intensity, dtype=float)
if bridge_dropouts > 0:
xic = XIC(target_mz=xic.target_mz, rt=np.asarray(xic.rt, dtype=float),
intensity=_bridge_gaps(raw, int(bridge_dropouts)))
b = peak_boundary(xic, contains_rt=contains_rt, apex_window=apex_window, threshold=threshold,
baseline_threshold=baseline_threshold,
baseline_quantile=baseline_quantile,
Expand Down Expand Up @@ -720,7 +755,7 @@ def peak_metrics(xic: "XIC", *,
snr = apex_int / mad if mad > 0 else float("nan")
q10 = float(np.quantile(inten, baseline_quantile))
prominence = (apex_int - q10) / q10 if q10 > 0 else float("nan")
n_signal = int(np.count_nonzero(seg_int > q10))
n_signal = int(np.count_nonzero(raw[left:right + 1] > q10))

beta_cor, beta_snr = beta_values(seg_int, seg_rt) if len(seg_int) >= 5 else (float("nan"),) * 2

Expand Down Expand Up @@ -812,12 +847,7 @@ def find_peaks(xic: "XIC", *,
if n < 3 or max_peaks < 1:
return []
raw = inten
if bridge_dropouts > 0:
inten = inten.copy()
nz = np.flatnonzero(raw > 0)
for a, b in zip(nz[:-1], nz[1:]):
if 1 < b - a <= bridge_dropouts + 1:
inten[a + 1:b] = np.interp(np.arange(a + 1, b), [a, b], [raw[a], raw[b]])
inten = _bridge_gaps(raw, int(bridge_dropouts))
baseline = float(np.quantile(inten, baseline_quantile)) # as peak_boundary
prev_ok = np.r_[True, inten[1:] >= inten[:-1]]
next_ok = np.r_[inten[:-1] >= inten[1:], True]
Expand Down
20 changes: 20 additions & 0 deletions tests/test_xic.py
Original file line number Diff line number Diff line change
Expand Up @@ -537,6 +537,26 @@ def test_bridge_dropouts_rejoins_the_split_elution(self):
assert len(two) == 1
assert y[52] == 0.0 # the caller's trace is not modified

def test_bridge_dropouts_helper(self):
from mzml_utils import bridge_dropouts
y = np.array([0, 4, 0, 8, 0, 0, 2, 0], dtype=float)
assert bridge_dropouts(y, 1).tolist() == [0, 4, 6, 8, 0, 0, 2, 0]
assert bridge_dropouts(y, 2).tolist() == [0, 4, 6, 8, 6, 4, 2, 0]
assert bridge_dropouts(y, 0).tolist() == y.tolist() and y[2] == 0 # input untouched

def test_peak_metrics_bridged_agrees_with_find_peaks(self):
from mzml_utils.xic import find_peaks
rt = np.linspace(0, 10, 101)
y = np.where(np.abs(rt - 5) < 1.5, _gauss(rt, 5, 0.5, 1e5), 0.0)
y[53] = 0.0 # a dropout just after the apex
xic = XIC(0.0, rt, y)
cut = peak_metrics(xic, contains_rt=5.0, nearest_valley=True)
assert cut.right_rt == pytest.approx(5.3) # unbridged: the dropout ends the peak
pm = peak_metrics(xic, contains_rt=5.0, nearest_valley=True, bridge_dropouts=1)
(p,) = find_peaks(xic, bridge_dropouts=1)
assert (pm.left_rt, pm.apex_rt, pm.right_rt) == (p.left_rt, p.apex_rt, p.right_rt)
assert pm.n_signal == p.n_signal == pm.n_points - 3 # two empty boundaries + the dropout

def test_n_signal_counts_only_scans_with_the_ion(self):
from mzml_utils.xic import find_peaks
rt = np.arange(11, dtype=float)
Expand Down
Loading