exprmat 0.1.37__tar.gz → 0.1.38__tar.gz

{exprmat-0.1.37 → exprmat-0.1.38}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: exprmat
-Version: 0.1.37
+Version: 0.1.38
 Classifier: Programming Language :: Rust
 Classifier: Programming Language :: Python :: Implementation :: CPython
 Classifier: Programming Language :: Python :: Implementation :: PyPy
@@ -55,7 +55,6 @@ Requires-Dist: hdbscan
 Requires-Dist: graphtools
 Requires-Dist: bincsync>=0.3.0
 Requires-Dist: pybigwig
-Requires-Dist: pybedgraph
 Requires-Dist: loompy
 Requires-Dist: squidpy
 Requires-Dist: zarr

{exprmat-0.1.37 → exprmat-0.1.38}/exprmat/__init__.py

@@ -128,6 +128,26 @@ if PERFORM_DATABASE_CHECK:
         warning(f'db version: {db_ver}  package version: {importlib.metadata.version("exprmat")}')
 
 
+def cuda():
+
+    import torch
+    info(f'PyTorch version: {torch.__version__}')
+    info(f'GPU acceleration availability: {"YES" if torch.cuda.is_available() else "NO"}')
+    if torch.cuda.is_available():
+        info(f'CUDA version: {torch.version.cuda}')
+        info(f'Number of installed GPUs: {torch.cuda.device_count()}')
+        info(f'Supporting BF16 format: {"YES" if torch.cuda.is_bf16_supported() else "NO"}')
+        info(f'Devices:')
+        for i in range(torch.cuda.device_count()):
+            total_mem = torch.cuda.get_device_properties(i).total_memory / 1024 / 1024 / 1024
+            dedicated_mem = torch.cuda.memory_allocated(i) / 1024 / 1024 / 1024
+            info(f'[{i}] {torch.cuda.get_device_name(i)} {"*" if torch.cuda.current_device() == i else ""}')
+            info(f'    CUDA capability: {torch.cuda.get_device_capability(i)}')
+            info(f'    Installed VRAM (GiB): {total_mem:.2f} GiB')
+            info(f'    Supporting TensorCore: {"YES" if torch.cuda.get_device_properties(0).major >= 7 else "NO"}')
+            info(f'    Current dedicated memory: {dedicated_mem:.2f} / {total_mem:.2f} GiB ({100 * dedicated_mem / total_mem:.1f}%)')
+
+
 def version(): 
     import os
     import sys

{exprmat-0.1.37 → exprmat-0.1.38}/exprmat/data/presets.py

@@ -40,5 +40,40 @@ g2m_genesets = {
 
 preset_genesets = {
     's': s_gene_sets,
-    'g2m': g2m_genesets
+    'g2m': g2m_genesets,
+    'custom': {
+        'mmu': {
+            't': [
+                'Cd3d', 'Cd8a', 'Cd4', 'Sell', 
+                'Il7r', 'Tcf7', 'Ifng', 'Gzmb', 
+                'Tox', 'Pdcd1', 'Havcr2', 'Lag3',
+                'Mki67', 'Icos', 'Foxp3', 'Id2',
+                'Id3', 'Gata3', 'Tbx21', 'Cd69',
+                'Cxcr5', 'Bach1', 'Cxcr5', 'Xcl1',
+                'Etv6', 'Ikzf3', 'Zbed2', 'Batf',
+                'Klrb1c', 'Klrg1', 'Prf1'
+            ],
+            'chemokines': {
+                "CCL": [
+                    "Ccl1", "Ccr1l1", "Ccl2", "Ccl3", "Ccl4", "Ccl5", "Ccl6",
+                    "Ccl7", "Ccl17", "Ccl8", "Ccl9", "Ccl11", "Ccl12", "Ccl19",
+                    "Ccl19-ps2", "Ccl20", "Ccl22", "Ccl24", "Ccl25", "Ccl26",
+                    "Ccl27a", "Ccl28",
+                ],
+                "CCR": [
+                    "Ccr1", "Ccr2", "Ccr3", "Ccr4", "Ccr5", "Ccr6",
+                    "Ccr7", "Ccr8", "Ccr9", "Ccr10",
+                ],
+                "CXCL": [
+                    "Cxcl1", "Cxcl2", "Cxcl3", "Cxcl5", "Cxcl9", "Cxcl10", "Cxcl11",
+                    "Cxcl12", "Cxcl13", "Cxcl14", "Cxcl15", "Cxcl16", "Cxcl17",
+                ],
+                "CXCR": ["Cxcr1", "Cxcr2", "Cxcr3", "Cxcr4", "Cxcr5", "Cxcr6"],
+                "XCL": ["Xcl1"],
+                "XCR": ["Xcr1"],
+                "CX3CL": ["Cx3cl1"],
+                "CX3CR": ["Cx3cr1"],
+            }
+        }
+    }
 }

exprmat-0.1.38/exprmat/peaks/linkage.py

@@ -0,0 +1,19 @@
+
+# Peak-to-gene linkage
+# --------------------
+# 
+# We estimated a linkage score for each peak-gene pair using linear regression models, 
+# based on recent work described in the SHARE-seq method [Ma et al., 2020]. For 
+# each gene, we computed the Pearson correlation coefficient r between the gene 
+# expression and the accessibility of each peak within 500 kb of the gene TSS. 
+# For each peak, we then computed a background set of expected correlation coefficients 
+# given properties of the peak by randomly sampling 200 peaks located on a different 
+# chromosome to the gene, matched for GC content, accessibility, and sequence length 
+# (MatchRegionStats function in Signac). We then computed the Pearson correlation 
+# between the expression of the gene and the set of background peaks. A z-score 
+# was computed for each peak as z = (r − µ)/σ, where µ was the background mean 
+# correlation coefficient and σ was the standard deviation of the background 
+# correlation coefficients for the peak. We computed a p-value for each peak using 
+# a one-sided z-test, and retained peak-gene links with a p-value < 0.05 and a 
+# Pearson correlation coefficient > 0.05 or < -0.05. This was performed using the 
+# LinkPeaks function in Signac.

exprmat-0.1.38/exprmat/peaks/motif.py

@@ -0,0 +1,54 @@
+
+# Motif enrichment
+# ----------------
+#
+# A hypergeometric test was used to test for overrepresentation of each DNA 
+# motif in the set of differentially accessible peaks compared to a background 
+# set of peaks. We tested motifs present in the JASPAR database [Fornes et al., 2020] 
+# for human (species code 9606) by first identifying which peaks contained each 
+# motif using the motifmatchr R package [Schep, 2020]. We computed the GC 
+# content (percentage of G and C nucleotides) for each differentially accessible 
+# peak and sampled a background set of 40,000 peaks such that the background set 
+# was matched for overall GC content, accessibility, and peak width. This was 
+# performed using the FindMotifs function in Signac.
+# 
+# Motif footprinting
+# ------------------
+#
+# We performed transcription factor motif footprinting following previously 
+# described methods [Corces et al., 2018]. To account for Tn5 sequence insertion 
+# bias, we first computed the observed Tn5 insertion frequency at each DNA 
+# hexamer using all Tn5 insertions on chromosome 1. This was done by extracting 
+# the base-resolution Tn5 insertion positions for each fragment mapped to 
+# chromosome 1, and extending the insertion coordinate 3 bp upstream and 2 bp 
+# downstream. We then extracted the DNA sequence corresponding to these 
+# coordinates using the getSeq function from the Biostrings R package 
+# [Pagès et al., 2020] and counted the frequency of each hexamer using the table 
+# function in R. We next computed the expected Tn5 hexamer insertion frequencies 
+# based on the frequency of each hexamer on chromosome 1. We counted the 
+# frequency of each hexamer using the oligonucleotideFrequency function in the 
+# Biostrings package with width=6 and names=“chr1”, using the hg38 genome via 
+# the BSgenome R package [Pagès, 2020]. Finally, we computed the Tn5 insertion 
+# bias as the observed Tn5 insertions divided by the expected insertions at each 
+# hexamer. This was performed using the InsertionBias function in Signac.
+# 
+# To perform motif footprinting, we first identified the coordinates of each 
+# instance of the motif to be footprinted using the matchMotifs function from 
+# the motifmatchr package with out=“positions” to return the genomic coordinates 
+# of each motif instance [Schep, 2020]. Motif coordinates were then resized to 
+# include the +/-250 bp sequence. The Tn5 insertion frequency was counted at each 
+# position in the region for each motif instance to produce a matrix containing 
+# the total observed Tn5 insertion events at each position relative to the motif 
+# center for each cell. We then found the expected Tn5 insertion frequency matrix 
+# by computing the hexamer frequency matrix, M. The hexamer frequency matrix M 
+# was defined as a matrix with i rows corresponding to i different DNA hexamers 
+# and j columns corresponding to j positions centered on the motif, and each 
+# entry Mi j corresponded to the hexamer count for hexamer i at position j. To 
+# find the expected Tn5 insertion frequency at each position relative to the motif 
+# given the Tn5 insertion bias (see above), we computed the matrix cross product 
+# between the hexamer frequency matrix M and the Tn5 insertion bias vector. 
+# Finally, the expected Tn5 insertion frequencies were normalized by dividing by 
+# the mean expected frequency in the 50 bp flanking regions (the regions 200 to 
+# 250 bp from the motif). To correct for Tn5 insertion bias we subtracted the 
+# expected Tn5 insertion frequencies from the observed Tn5 insertion frequencies 
+# at each position. This was performed using the Footprint function in Signac.

{exprmat-0.1.37 → exprmat-0.1.38}/exprmat/reader/static.py

@@ -805,7 +805,6 @@ def atac_infer_gene_activity(adata, sample_name, make_gene_args = {}, exact = Tr
         gmodel = gmodel.loc[gmodel['type'] == 'gene', :].copy()
         bedgraphs = adata.obs['bedgraph'].tolist()
         import pyBigWig as pybw
-        import pyBedGraph as pybdg
         from rich.progress import track
         
         # generate a matrix with n_peaks * n_samples
@@ -819,7 +818,6 @@ def atac_infer_gene_activity(adata, sample_name, make_gene_args = {}, exact = Tr
             if os.path.isfile(fp.replace('.bdg', '.bigwig')):
                 fhandle = pybw.open(fp.replace('.bdg', '.bigwig'))
                 ftype = 'bigwig'
-            else: fhandle = pybdg.BedGraph('', fp)
             
             for i_region in track(range(len(gmodel)), description = f'querying genes for {fp}'):
 
@@ -976,7 +974,6 @@ def atac_make_peak_matrix(adata, sample_name, key_peaks = 'peaks.merged', exact
         bedgraphs = adata.obs['bedgraph'].tolist()
         
         import pyBigWig as pybw
-        import pyBedGraph as pybdg
         from rich.progress import track
         
         # generate a matrix with n_peaks * n_samples
@@ -990,7 +987,6 @@ def atac_make_peak_matrix(adata, sample_name, key_peaks = 'peaks.merged', exact
             if os.path.isfile(fp.replace('.bdg', '.bigwig')):
                 fhandle = pybw.open(fp.replace('.bdg', '.bigwig'))
                 ftype = 'bigwig'
-            else: fhandle = pybdg.BedGraph('', fp)
             
             for i_region in track(range(len(peak_table)), description = f'querying peaks for {fp}'):
 

{exprmat-0.1.37 → exprmat-0.1.38}/pyproject.toml

@@ -20,7 +20,7 @@ description = "exprmat: Routines for expression matrices"
 readme = "readme.md"
 license = {file = "license"}
 requires-python = ">=3.9, <3.13"
-version = "0.1.37"
+version = "0.1.38"
 authors = [
     {name = "Zheng Yang", email = "xornent@outlook.com"}
 ]
@@ -83,7 +83,6 @@ dependencies = [
     'graphtools',
     'bincsync >= 0.3.0',
     'pyBigWig',
-    'pyBedGraph',
     'loompy',
     'squidpy',
     'zarr',

exprmat-0.1.37/exprmat/data/delim.py

@@ -1,89 +0,0 @@
-
-# utilities of reading delimited files from raw or gzipped archives.
-
-from exprmat.ansi import error
-
-
-def readdelim(fname, encoding = 'utf-8', 
-              delim = '\t', ignore_first_column = True):
-    
-    with open(fname, 'r', encoding = encoding) as f:
-        return readdelim_fp(f, delim, ignore_first_column)
-    
-    pass
-
-
-def readdelim_gzipped(fname, encoding = 'utf-8', 
-                      delim = '\t', ignore_first_column = True):
-
-    import gzip
-    import io
-
-    with gzip.open(fname, 'rb') as f:
-        with io.TextIOWrapper(f, encoding = encoding) as enc:
-            return readdelim_fp(enc, delim, ignore_first_column)
-    
-    pass
-
-
-def readdelim_fp(fp, delim = '\t', ignore_first_column = True):
-    
-    line = fp.readline().replace('\n', '')
-    print('', end = '')
-    
-    header = False
-    header_cols = []
-
-    lines = { }
-    lineno = 0
-    rowno = 0
-
-    while line:
-
-        # the file has been scanned by scangff. so there is no format problems.
-        # we do not perform any checks.
-
-        if len(line) == 0:
-            line = fp.readline().replace('\n', '')
-            lineno += 1; continue
-        
-        elif line[0] == '#':
-            line = fp.readline().replace('\n', '')
-            lineno += 1; continue
-        
-        else: rowno += 1
-
-        if not header:
-
-            cols = line.split(delim)
-            if len(cols) > 1:
-                if ignore_first_column:
-                    for x in cols[1:]: lines[x] = []
-                    header_cols = cols[1:]
-                else:
-                    for x in cols: lines[x] = []
-                    header_cols = cols
-            else:
-                if ignore_first_column: return {}
-                else: lines[cols[0]] = []; header_cols = cols
-
-            header = True
-            line = fp.readline().replace('\n', '')
-            lineno += 1; continue
-        
-        cols = line.split(delim)
-        if ignore_first_column: cols = cols[1:]
-
-        i = 0
-        for c in header_cols:
-            lines[c] += [cols[i]]
-            i += 1
-
-        line = fp.readline().replace('\n', '')
-        lineno += 1
-
-        if rowno % 1000 == 0:
-            print('\rreaddelim_fp(fp) progress:', rowno, end = '')
-    
-    print('')
-    return lines

exprmat-0.1.37/exprmat/data/gff.py

@@ -1,162 +0,0 @@
-
-# utilities of reading gff files from raw or gzipped archives.
-
-from exprmat.ansi import error
-
-
-def readgff(fname, encoding = 'utf-8'):
-    
-    with open(fname, 'r', encoding = encoding) as f:
-        col, _, _ = scangff(f)
-        return readgff_fp(f, col)
-    
-    pass
-
-
-def readgff_gzipped(fname, encoding = 'utf-8'):
-
-    import gzip
-    import io
-
-    with gzip.open(fname, 'rb') as f:
-        with io.TextIOWrapper(f, encoding = encoding) as enc:
-            col, _, _ = scangff(enc)
-            return readgff_fp(enc, col)
-    
-    pass
-
-
-def gffcols(fname, encoding = 'utf-8'):
-    
-    with open(fname, 'r', encoding = encoding) as f:
-        col, _, type = scangff(f)
-    
-    return col, type
-
-
-def gffcols_gzipped(fname, encoding = 'utf-8'):
-
-    import gzip
-    import io
-
-    with gzip.open(fname, 'rb') as f:
-        with io.TextIOWrapper(f, encoding = encoding) as enc:
-            col, _, type = scangff(enc)
-    
-    return col, type
-
-
-# scan the gff table, and get the full union of available columns (though some
-# may not present in certain kind of rows.) this help later specify the classes
-# and columns of a full union table, which contains all of the gff metadata
-# column information. also returns the available data row numbers.
-
-def scangff(fp):
-
-    line = fp.readline().replace('\n', '')
-    lineno = 0
-    rowno = 0
-    cols = ['.seqid', '.source', '.type', '.start', '.end', '.score', '.strand', '.phase']
-    attrcols = []
-    typecols = {}
-    print('', end = '')
-
-    while line:
-        
-        if len(line) == 0:
-            line = fp.readline().replace('\n', '')
-            lineno += 1; continue
-        
-        elif line[0] == '#':
-            line = fp.readline().replace('\n', '')
-            lineno += 1; continue
-        
-        else: rowno += 1
-
-        tabdelim = line.split('\t')
-        if len(tabdelim) != 9: 
-            print('')
-            error('line {0} does not have exactly 9 columns!'.format(lineno))
-        
-        if len(tabdelim[8]) > 0:
-            attribs = tabdelim[8].split(';')
-            rnames = [x.split('=')[0] for x in attribs]
-            
-            typename = tabdelim[2]
-            if not typename in typecols.keys(): typecols[typename] = { '.': 1 }
-            else: typecols[typename]['.'] += 1
-
-            for rname in rnames:
-                if not rname in attrcols: attrcols += [rname]
-
-                if not rname in typecols[typename].keys(): typecols[typename][rname] = 1
-                else: typecols[typename][rname] += 1
-
-        line = fp.readline().replace('\n', '')
-        lineno += 1
-
-        if rowno % 1000 == 0:
-            print('\rscangff(fp) progress:', rowno, end = '')
-    
-    fp.seek(0) # seek back to the origin.
-    print('')
-    
-    return cols + attrcols, rowno, typecols
-
-
-def readgff_fp(fp, columns):
-    
-    line = fp.readline().replace('\n', '')
-    print('', end = '')
-
-    lines = { }
-    for i in columns: lines[i] = []
-    lineno = 0
-    rowno = 0
-
-    while line:
-
-        # the file has been scanned by scangff. so there is no format problems.
-        # we do not perform any checks.
-
-        if len(line) == 0:
-            line = fp.readline().replace('\n', '')
-            lineno += 1; continue
-        
-        elif line[0] == '#':
-            line = fp.readline().replace('\n', '')
-            lineno += 1; continue
-        
-        else: rowno += 1
-
-        # ['.seqid', '.source', '.type', '.start', '.end', '.score', '.strand', '.phase']
-        
-        tabdelim = line.split('\t')
-        lines['.seqid'] += [tabdelim[0]]
-        lines['.source'] += [tabdelim[1]]
-        lines['.type'] += [tabdelim[2]]
-        lines['.start'] += [tabdelim[3]]
-        lines['.end'] += [tabdelim[4]]
-        lines['.score'] += [tabdelim[5]]
-        lines['.strand'] += [tabdelim[6]]
-        lines['.phase'] += [tabdelim[7]]
-
-        attl = columns[8:]
-        if len(tabdelim[8]) > 0:
-            attribs = tabdelim[8].split(';')
-            for attrib in attribs:
-                namevalue = attrib.split('=')
-                lines[namevalue[0]] += [namevalue[1]]
-                attl.remove(namevalue[0])
-        
-        for a in attl:
-            lines[a] += [None]
-
-        line = fp.readline().replace('\n', '')
-        lineno += 1
-
-        if rowno % 1000 == 0:
-            print('\rreadgff_fp(fp, columns) progress:', rowno, end = '')
-    
-    print('')
-    return lines

exprmat-0.1.37/exprmat/data/private.py

@@ -1,70 +0,0 @@
-
-mmu_chemokines = {
-    "CCL": [
-        "Ccl1",
-        "Ccr1l1",
-        "Ccl2",
-        "Ccl3",
-        "Ccl4",
-        "Ccl5",
-        "Ccl6",
-        "Ccl7",
-        "Ccl17",
-        "Ccl8",
-        "Ccl9",
-        "Ccl11",
-        "Ccl12",
-        "Ccl19",
-        "Ccl19-ps2",
-        "Ccl20",
-        "Ccl22",
-        "Ccl24",
-        "Ccl25",
-        "Ccl26",
-        "Ccl27a",
-        "Ccl28",
-    ],
-    "CCR": [
-        "Ccr1",
-        "Ccr2",
-        "Ccr3",
-        "Ccr4",
-        "Ccr5",
-        "Ccr6",
-        "Ccr7",
-        "Ccr8",
-        "Ccr9",
-        "Ccr10",
-    ],
-    "CXCL": [
-        "Cxcl1",
-        "Cxcl2",
-        "Cxcl3",
-        "Cxcl5",
-        "Cxcl9",
-        "Cxcl10",
-        "Cxcl11",
-        "Cxcl12",
-        "Cxcl13",
-        "Cxcl14",
-        "Cxcl15",
-        "Cxcl16",
-        "Cxcl17",
-    ],
-    "CXCR": ["Cxcr1", "Cxcr2", "Cxcr3", "Cxcr4", "Cxcr5", "Cxcr6"],
-    "XCL": ["Xcl1"],
-    "XCR": ["Xcr1"],
-    "CX3CL": ["Cx3cl1"],
-    "CX3CR": ["Cx3cr1"],
-}
-
-mmu_t_signatures = {
-    'Cd3d', 'Cd8a', 'Cd4', 'Sell', 
-    'Il7r', 'Tcf7', 'Ifng', 'Gzmb', 
-    'Tox', 'Pdcd1', 'Havcr2', 'Lag3',
-    'Mki67', 'Icos', 'Foxp3', 'Id2',
-    'Id3', 'Gata3', 'Tbx21', 'Cd69',
-    'Cxcr5', 'Bach1', 'Cxcr5', 'Xcl1',
-    'Etv6', 'Ikzf3', 'Zbed2', 'Batf',
-    'Klrb1c', 'Klrg1', 'Prf1'
-}