Merge pull request #36 from sfiligoi/multi_230320

Add multi mode

.github/workflows/main.yml

@@ -154,6 +154,10 @@ jobs:
         python compare_unifrac_stats.py t1.h5 5 999 1.001112 0.456  0.001 0.1
         ls -l t1.h5
         rm -f t1.h5
+        time ssu -m unweighted -i test500.biom  -t test500.tre --pcoa 4  -r hdf5 --permanova 99 -g test500.tsv -c empo_2 -o t1.h5
+        python compare_unifrac_stats.py t1.h5 5 99 1.001112 0.456  0.001 0.2
+        ls -l t1.h5
+        rm -f t1.h5
         time ssu -m weighted_unnormalized_fp32 -i test500.biom  -t test500.tre --pcoa 4  -r hdf5_nodist -g test500.tsv -c empo_3 -o t1.h5
         # compare to values given by skbio.stats.distance.permanova
         python compare_unifrac_stats.py t1.h5 17 999 0.890697 0.865 0.001 0.1
@@ -181,6 +185,25 @@ jobs:
         ls -l t1.h5
         rm -f t1.h5
         rm -f t1.partial.*
+        # subsample
+        echo "subsample"
+        time ssu -f -m unweighted -i test500.biom  -t test500.tre --pcoa 4  -r hdf5_fp32 --subsample-depth 100 -o t1.h5
+        ./compare_unifrac_pcoa.py test500.unweighted_fp32.f.h5 t1.h5 3 0.3
+        rm -f t1.h5
+        time ssu -m unweighted -i test500.biom  -t test500.tre --pcoa 4  -r hdf5_nodist --permanova 99 -g test500.tsv -c empo_2 --subsample-depth 100 -o t1.h5
+        python compare_unifrac_stats.py t1.h5 5 99 1.001112 0.456  0.05 0.5
+        ls -l t1.h5
+        rm -f t1.h5
+        # multi
+        echo "multi"
+        time ssu -f -m unweighted -i test500.biom  -t test500.tre --pcoa 4 --mode multi --subsample-depth 100 --n-subsamples 10 -o t1.h5
+        ./compare_unifrac_pcoa_multi.py test500.unweighted_fp32.f.h5 t1.h5 10 3 0.3
+        ls -l t1.h5
+        rm -f t1.h5
+        time ssu -m unweighted -i test500.biom  -t test500.tre --pcoa 4  --mode multi --n-subsamples 10 --permanova 99 -g test500.tsv -c empo_2 --subsample-depth 100 -o t1.h5
+        python compare_unifrac_stats_multi.py t1.h5 10 5 99 1.001112 0.456  0.08 0.5
+        ls -l t1.h5
+        rm -f t1.h5
         popd
     - name: Sanity checks
       shell: bash -l {0}

README.md

@@ -148,19 +148,21 @@ The methods can be used directly through the command line after install:
                                 partial : Compute UniFrac over a subset of stripes.
                                 partial-report : Start and stop suggestions for partial compute.
                                 merge-partial : Merge partial UniFrac results.
+                                multi : compute UniFrac multiple times.
         --start	[OPTIONAL] If mode==partial, the starting stripe.
         --stop	[OPTIONAL] If mode==partial, the stopping stripe.
         --partial-pattern	[OPTIONAL] If mode==merge-partial, a glob pattern for partial outputs to merge.
         --n-partials 	[OPTIONAL] If mode==partial-report, the number of partitions to compute.
         --report-bare	[OPTIONAL] If mode==partial-report, produce barebones output.
         --n-substeps 	[OPTIONAL] Internally split the problem in n substeps for reduced memory footprint, default is 1.
         --format|-r	[OPTIONAL]  Output format:
-                                 ascii : [DEFAULT] Original ASCII format.
+                                 ascii : Original ASCII format. (default if mode==one-off)
+                                 hdf5_nodist : HFD5 format, no distance matrix. (default if mode==multi)
                                  hdf5 : HFD5 format.  May be fp32 or fp64, depending on method.
                                  hdf5_fp32 : HFD5 format, using fp32 precision.
                                  hdf5_fp64 : HFD5 format, using fp64 precision.
-                                 hdf5_nodist : HFD5 format, no distance matrix, just PCoA.
-        --subsample-depth [OPTIONAL] Depth of subsampling of the input BIOM before computing unifrac
+        --subsample-depth   Depth of subsampling of the input BIOM before computing unifrac (required for mode==multi, optional for one-off)
+        --n-subsamples	[OPTIONAL] if mode==multi, number of subsampled UniFracs to compute (default: 100)
         --permanova	[OPTIONAL] Number of PERMANOVA permutations to compute (default: 999 with -g, do not compute if 0)
         --pcoa	[OPTIONAL] Number of PCoA dimensions to compute (default: 10, do not compute if 0)
         --seed	[OPTIONAL] Seed to use for initializing the random gnerator