Fix bug in simulation of relate datasets where some reads could have …

…zero coverage; enable _check_no_coverage_reads to handle masked arrays where the mask is a boolean value

src/seismicrna/core/batch/ends.py

@@ -74,7 +74,7 @@ def _check_no_coverage_reads(seg_ends: np.ndarray, what: str = "seg_ends"):
     num_reads, num_segs = count_reads_segments(seg_ends, what)
     if np.ma.isarray(seg_ends):
         # Check if any reads have no coverage (are completely masked).
-        no_coverage = seg_ends.mask.all(axis=1)
+        no_coverage = np.ma.getmaskarray(seg_ends).all(axis=1)
         if np.any(no_coverage):
             raise ValueError(
                 f"Got {np.count_nonzero(no_coverage)} read(s) with no coverage "

src/seismicrna/core/version.py

@@ -7,7 +7,7 @@
 
 logger = getLogger(__name__)
 
-__version__ = "0.17.3"
+__version__ = "0.17.4"
 
 
 def parse_version(version: str = __version__):

src/seismicrna/relate/batch.py

@@ -104,6 +104,10 @@ def simulate(cls,
                                                      (read_length if paired
                                                       else 0),
                                                      p_rev)
+        # Drop any reads with zero coverage.
+        has_coverage = np.less_equal(seg_end5s, seg_end3s).any(axis=1)
+        seg_end5s = seg_end5s[has_coverage]
+        seg_end3s = seg_end3s[has_coverage]
         simulated_all = cls(batch=batch,
                             section=section,
                             seg_end5s=seg_end5s,

src/seismicrna/relate/sim.py

@@ -156,6 +156,8 @@ def simulate_relate(out_dir: Path,
                               phred_enc=0,
                               min_phred=0,
                               ambindel=False,
+                              clip_end5=0,
+                              clip_end3=0,
                               min_reads=num_reads,
                               n_reads_xam=num_reads,
                               n_reads_rel=num_reads,