Explore the tables from Python

Everything the window shows is available from Python. A Context is one organism's gene table, its measured networks and the leakage guard every analysis applies.

In [1]:

from starplast import strategies as Sctx = S.shipped('Tg')print(ctx.organism, ctx.n, 'genes,', len(ctx.nodes.columns), 'columns')

Out[1]:

Tg 8140 genes, 438 columns

In [2]:

hits = ctx.nodes[ctx.nodes['product'].str.contains('microneme protein', na=False)]hits[['gene_id', 'product', 'compartment', 'fit_invitro_hff']].head(8)

Out[2]:

gene_id	product	compartment	fit_invitro_hff
TGME49_267680	microneme protein MIC12	micronemes	0.5
TGME49_201780	microneme protein MIC2	micronemes	-1.17
TGME49_245490	microneme protein MIC8	micronemes	-1.71
TGME49_289630	microneme protein MIC16	micronemes	1.98
TGME49_254430	microneme protein, putative	unassigned	-0.27
TGME49_200230	microneme protein MIC17C	unassigned	NaN
TGME49_208030	microneme protein MIC4	micronemes	0.83
TGME49_222188	microneme protein MIC21L1	unassigned	-0.8

In [3]:

ctx.truth('compartment').value_counts().head(10)

Out[3]:

	count
compartment
nucleus - chromatin	769
cytosol	599
nucleus - non-chromatin	461
mitochondrion - soluble	274
ER	248
dense granules	193
mitochondrion - membranes	193
apicoplast	167
Golgi	153
PM - integral	135

In [4]:

print(len(ctx.categorical_columns()), 'labels can be held out, e.g.', ctx.categorical_columns()[:6])print(len(ctx.numeric_columns()), 'measurements are available as features')

Out[4]:

18 labels can be held out, e.g. ['dtm_class', 'lopit_map', 'lopit_mcmc', 'compartment', 'lopit_unified', 'ortholopit_label']
367 measurements are available as features

What a held-out analysis never sees

Holding a label out removes the label, anything that restates it (measured association of 0.8 or more, on values or ranks), the experiment that produced it, the same quantity measured another way, and any edge layer built from it. This is what makes a recovered label mean something.

In [5]:

sorted(ctx.banned('compartment'))

Out[5]:

['compartment', 'compartment_best', 'compartment_source', 'cyst_wall_max_spectral', 'cyst_wall_n_baits', 'lopit_confident', 'lopit_map', 'lopit_mcmc', 'lopit_methods_agree', 'lopit_prob_map', 'lopit_prob_mcmc', 'lopit_unified', 'ortholopit_accepted', 'ortholopit_accuracy', 'ortholopit_donors', 'ortholopit_label', 'pvm_proximity_positive', 'secretome_soluble_over_vesicle_log2']

In [6]:

print('edge layers:', ctx.layers())print('banned when compartment is held out:', sorted(ctx.banned_layers('compartment')))

Out[6]:

edge layers: ['coexpression', 'cofitness', 'comention', 'comention_ft', 'compartment', 'cotranslation', 'domain', 'ip_ms', 'orthogroup', 'struct', 'structural_hole', 'unwritten_interaction', 'xlms']
banned when compartment is held out: ['compartment']