CPP.run_num

CPP.run_num(dict_num_parts, labels, label_test=1, label_ref=0, n_filter=100, n_pre_filter=None, pct_pre_filter=5, max_std_test=0.2, max_overlap=0.5, max_cor=0.5, check_cat=True, parametric=False, start=1, tmd_len=20, jmd_n_len=10, jmd_c_len=10, n_jobs=None, vectorized=True, n_batches=None, n_sample_batches=None, return_stats=False, redundancy='legacy')[source]

Numerical-mode Comparative Physicochemical Profiling (CPP): same algorithm as run(), but per-residue values come from a pre-sliced numerical tensor (dict_num_parts) instead of an AA→scale lookup. Use for PLM embeddings, DSSP one-hots, PTM dummies, or any per-residue numerical representation.

Same pipeline (pre-filter stats, pre-filter, recompute, add_stat, redundancy filter) and same output schema as run(). The constructor-bound df_scales / df_cat provide DIMENSION NAMES + categories for the D axis of dict_num_parts (the per-AA values they would normally provide are unused — dict_num_parts is the value source).

Added in version 1.1.0.

Changed in version 1.1.0: Honors bootstrap stability annotation (CPP(bootstrap=True)) exactly like run(), resampling along the sample axis of dict_num_parts and adding a selection_frequency column to df_feat (the selected features are unchanged).

Parameters:

dict_num_parts (dict[str, np.ndarray], required) – Per-part NaN-padded numerical tensors, produced by NumericalFeature.get_parts(). Each value has shape (n_samples, L_part_max, D) aligned row-for-row with self.df_parts. Keys must match self.df_parts.columns. D must equal len(self.df_scales.columns) (each D dimension names a “scale”).
labels (array-like, shape (n_samples,)) – Class labels for samples in sequence DataFrame (typically, test=1, reference=0).
label_test (int, default=1) – Class label of test group in labels.
label_ref (int, default=0) – Class label of reference group in labels.
n_filter (int, default=100) – Number of features to be filtered/selected by CPP algorithm.
n_pre_filter (int, optional) – Number of features to be pre-filtered. If None, a percentage of all features is used.
pct_pre_filter (int, default=5) – Percentage of all features that should remain after the pre-filtering step.
max_std_test (float, default=0.2) – Maximum standard deviation [>0-<1] within the test group used as threshold for pre-filtering.
max_overlap (float, default=0.5) – Maximum positional overlap [0-1] of features used as threshold for filtering.
max_cor (float, default=0.5) – Maximum Pearson correlation [0-1] of feature scales used as threshold for filtering.
check_cat (bool, default=True) – Whether to check for redundancy within scale categories during filtering.
redundancy ({'legacy', 'exact'}, default='legacy') – Position-overlap criterion of the redundancy-reduction step. 'legacy' (default) keeps results reproducible across versions; 'exact' compares the actual residue positions (see Notes).
parametric (bool, default=False) – Whether to use parametric (T-test) or non-parametric (Mann-Whitney U test) for p-value computation.
start (int, default=1) – Position label of first residue position (starting at N-terminus).
tmd_len (int, default=20) – Length of target middle domain (TMD) (>0).
jmd_n_len (int, default=10) – Length of JMD-N (>=0).
jmd_c_len (int, default=10) – Length of JMD-C (>=0).
n_jobs (int, None, or -1, default=None) – Number of CPU cores (>=1) used for multiprocessing. If None, the number is optimized automatically; if -1, all available cores are used. Overridden by options['n_jobs'] when set. The Python 3.14 + macOS spawn caveat documented in run() applies here too.
vectorized (bool, default=True) – Whether to apply sequence splitting and the Mann-Whitney U test in ‘vectorized’ mode (True), improving speed but increasing memory consumption. The vectorized Mann-Whitney U test uses a fast normal approximation of the p-value (roughly an order of magnitude faster on the test step); vectorized=False instead computes the exact scipy.stats.mannwhitneyu() p-value, which is slower but reproducible bit-for-bit. This choice changes only the reported ‘p_val_mann_whitney’ and ‘p_val_fdr_bh’ columns: feature ranking and selection are driven by ‘abs_auc’ and ‘abs_mean_dif’, so the selected features are identical in either mode.
n_batches (int, None, default=None) – Number of batches (2 to len(df_scales.columns)) over the D axis of dict_num_parts. If None, single-pass; a value bounds the pass-1 stat working set to one D-chunk (pass-2 recompute still runs globally, so this trims pass-1 memory but does not bound overall peak RSS — use n_sample_batches for that). Output is bit-exact with the single-pass result. Mutually exclusive with n_sample_batches.
n_sample_batches (int, None, default=None) – Number of batches (2 to n_samples) over the sample axis. If None, single-pass. A value processes ceil(n_samples / n_sample_batches) samples per batch, bounding the per-batch working set (stat intermediates + pass-2 recompute) to O(batch_size) — the lever that actually lowers peak RSS for large n (the resident input tensor itself is unchanged). Pass-1 std_test uses accumulator-style variance, so the result may differ from the single-pass run by ULP-level rounding (after the round(3) on the stat columns), which can reorder tie-broken features; hence opt-in, not the default. Mutually exclusive with n_batches.
return_stats (bool, default=False) – If True, also return the filter-funnel statistics (last_filter_stats_) as a second element (df_feat, stats); if False, return only df_feat.

Returns:

df_feat – Same schema as run().

Return type:

pd.DataFrame, shape (n_features, n_feature_info)

Raises:

ValueError – If dict_num_parts is None (use run() for seq-mode), or if its shape / part names / D don’t align with the constructor’s self.df_parts and self.df_scales.

Notes

redundancy behaves exactly as in run() (default 'legacy' keeps results reproducible; 'exact' compares the true residue positions — an optional enhancement, not a correctness fix).
Call order — ``get_parts`` then ``run_num``. dict_num_parts must come from NumericalFeature.get_parts() (step 1), which slices a raw df_seq + dict_num into the per-part tensors consumed here (step 2). There is no raw-df_seq / dict_num entry point on run_num; passing dict_num_parts=None raises (use run() for sequence-mode).
Peak memory is higher than :meth:`run` and scales with the input tensor. run_num carries the dense (n_samples, L_part_max, D) per-part tensor that dict_num_parts materializes, so its peak memory is roughly an order of magnitude above the same-data run() (which streams an AA→scale lookup and never builds that tensor) and grows with D; runtime is otherwise comparable to run(). For large n, pass n_sample_batches to bound the per-batch working set and lower peak RSS roughly in proportion to the batch count (e.g. n_sample_batches=10 cut working memory ~6× in internal benchmarks), at the cost of ULP-level non-determinism in tie-broken features. n_batches only trims the pass-1 stat working set, not overall peak.
Raw PLM embeddings are not directly usable — normalize them first. Per-residue values are expected in [0, 1] (the StructurePreprocessor / AnnotationPreprocessor normalization convention), since the default max_std_test=0.2 pre-filter is calibrated for that range. Raw embeddings (unbounded floats) must be passed through EmbeddingPreprocessor.encode() to obtain a [0, 1]-normalized {entry: (L, D)} dict_num before NumericalFeature.get_parts(). Skipping normalization raises no error — the max_std_test pre-filter is simply miscalibrated for the out-of-range spread, so the feature funnel silently keeps/drops the wrong features. (EmbeddingPreprocessor.build_scales / build_cat serve the other, AA-scale path via run(); they are not a per-residue value source here.)
Three arms, one entry point. structure-only (dict_num from StructurePreprocessor), embedding (EmbeddingPreprocessor.encode), and fused (concatenate sources with aaanalysis.combine_dict_nums() first) all flow through get_parts → run_num — only the dict_num differs.
Compositional vs positional features emerge from split_kws exactly as in run() (n_split_max=1 with no patterns ⇒ compositional whole-part mean; otherwise positional).

See also

run(): sequence-mode equivalent (no dict_num_parts).
NumericalFeature.get_parts(): produces (df_parts, dict_num_parts) from raw df_seq + dict_num.
EmbeddingPreprocessor, StructurePreprocessor, AnnotationPreprocessor: the per-residue dict_num sources (PLM embeddings / structure / annotations), combinable via aaanalysis.combine_dict_nums().

Examples

To demonstrate the CPP().run() method, we load the DOM_GSEC example dataset (see [Breimann25]):

import aaanalysis as aa
aa.options["verbose"] = False
df_seq = aa.load_dataset(name="DOM_GSEC")
labels = df_seq["label"].to_list()
sf = aa.SequenceFeature()
df_parts = sf.get_df_parts(df_seq=df_seq)

You just need to provide df_parts to the CPP object and run the algorithm with its respective labels using the CPP().run() method:

cpp = aa.CPP(df_parts=df_parts)
# Create >500,000 feature and filter them down to 100 features
df_feat = cpp.run(labels=labels)
aa.display_df(df_feat, n_rows=10, show_shape=True)

DataFrame shape: (100, 13)

	feature	category	subcategory	scale_name	scale_description	abs_auc	abs_mean_dif	mean_dif	std_test	std_ref	positions
1	TMD_C_JMD_C-Seg...2,3)-QIAN880106	Conformation	α-helix	α-helix (middle)	Weights for alp...ejnowski, 1988)	0.387000	0.118000	0.118000	0.068000	0.080000	27,28,29,30,31,32,33
2	TMD_C_JMD_C-Pat...,14)-CRAJ730103	Conformation	β-turn	β-turn	Normalized freq...d et al., 1973)	0.377000	0.285000	-0.285000	0.164000	0.177000	27,31
3	TMD_C_JMD_C-Seg...6,9)-FAUJ880104	Shape	Side chain length	Steric parameter	STERIMOL length...e et al., 1988)	0.367000	0.263000	0.263000	0.161000	0.168000	32,33
4	TMD_C_JMD_C-Seg...6,9)-ONEK900101	Others	Unclassified (Others)	ΔG values in peptides	Delta G values ...-DeGrado, 1990)	0.366000	0.111000	0.111000	0.070000	0.114000	32,33
5	TMD_C_JMD_C-Pat...,15)-QIAN880107	Conformation	α-helix	α-helix (middle)	Weights for alp...ejnowski, 1988)	0.363000	0.162000	0.162000	0.091000	0.118000	24,28,32,35
6	TMD_C_JMD_C-Seg...3,4)-HUTJ700103	Energy	Entropy	Entropy	Entropy of form...Hutchens, 1970)	0.360000	0.187000	0.187000	0.115000	0.128000	31,32,33,34,35
7	TMD_C_JMD_C-Seg...2,3)-WOLS870103	Others	PC 4	Principal Component 3 (Wold)	Principal prope...d et al., 1987)	0.359000	0.159000	-0.159000	0.090000	0.130000	27,28,29,30,31,32,33
8	TMD_C_JMD_C-Pat...,12)-CRAJ730103	Conformation	β-turn	β-turn	Normalized freq...d et al., 1973)	0.352000	0.227000	-0.227000	0.150000	0.170000	24,28,32
9	TMD_C_JMD_C-Seg...6,9)-MUNV940102	Energy	Free energy (folding)	Free energy (α-helix)	Free energy in ...-Serrano, 1994)	0.350000	0.129000	-0.129000	0.079000	0.124000	32,33
10	TMD_C_JMD_C-Seg...3,4)-WOLS870103	Others	PC 4	Principal Component 3 (Wold)	Principal prope...d et al., 1987)	0.341000	0.214000	-0.214000	0.128000	0.177000	31,32,33,34,35

Adjust Parts, Splits, and Scales as follows:

df_parts = sf.get_df_parts(df_seq=df_seq, list_parts=["tmd_jmd"])
split_kws = sf.get_split_kws(split_types=["Segment"], n_split_min=1, n_split_max=5)
# Load one of the provided top scale datasets
df_scales = aa.load_scales(top60_n=38)
# Create ~700 feature and filter them down to 19 features
cpp = aa.CPP(df_parts=df_parts, split_kws=split_kws, df_scales=df_scales)
df_feat = cpp.run(labels=labels)
aa.display_df(df_feat, n_rows=10, show_shape=True)

DataFrame shape: (19, 13)

	feature	category	subcategory	scale_name	scale_description	abs_auc	abs_mean_dif	mean_dif	std_test	std_ref	p_val_fdr_bh	positions
1	TMD_JMD-Segment...4,5)-ROBB760113	Conformation	β-turn	β-turn	Information mea...n-Suzuki, 1976)	0.316000	0.137000	-0.137000	0.102000	0.108000	0.000000	25,26,27,28,29,30,31,32
2	TMD_JMD-Segment...4,4)-ZIMJ680104	Energy	Isoelectric point	Isoelectric point	Isoelectric poi...n et al., 1968)	0.312000	0.099000	0.099000	0.069000	0.095000	0.000000	31,32,33,34,35,36,37,38,39,40
3	TMD_JMD-Segment...4,5)-KANM800103	Conformation	α-helix	α-helix	Average relativ...sa-Tsong, 1980)	0.297000	0.086000	0.086000	0.077000	0.068000	0.000000	25,26,27,28,29,30,31,32
4	TMD_JMD-Segment...5,5)-LINS030104	ASA/Volume	Accessible surface area (ASA)	ASA (folded protein)	Total median ac...s et al., 2003)	0.295000	0.141000	0.141000	0.115000	0.130000	0.000000	33,34,35,36,37,38,39,40
5	TMD_JMD-Segment...5,5)-JANJ780102	ASA/Volume	Buried	Buried	Percentage of b...n et al., 1978)	0.291000	0.130000	-0.130000	0.099000	0.124000	0.000000	33,34,35,36,37,38,39,40
6	TMD_JMD-Segment...5,5)-ZIMJ680103	Polarity	Hydrophilicity	Polarity (hydrophilicity)	Polarity (Zimme...n et al., 1968)	0.289000	0.178000	0.178000	0.159000	0.163000	0.000000	33,34,35,36,37,38,39,40
7	TMD_JMD-Segment...4,5)-FUKS010106	Composition	Membrane proteins (MPs)	Proteins of mesophiles (INT)	Interior compos...ishikawa, 2001)	0.277000	0.123000	0.123000	0.104000	0.127000	0.000000	25,26,27,28,29,30,31,32
8	TMD_JMD-Segment...4,4)-WOLR790101	Polarity	Hydrophobicity (surrounding)	Hydration potential	Hydrophobicity ...n et al., 1979)	0.267000	0.105000	-0.105000	0.100000	0.113000	0.000001	31,32,33,34,35,36,37,38,39,40
9	TMD_JMD-Segment...2,2)-CEDJ970105	Composition	AA composition	Nuclear proteins	Composition of ...o et al., 1997)	0.263000	0.062000	0.062000	0.062000	0.069000	0.000001	21,22,23,24,25,...,36,37,38,39,40
10	TMD_JMD-Segment...5,5)-MITS020101	Polarity	Amphiphilicity	Amphiphilicity	Amphiphilicity ...u et al., 2002)	0.262000	0.073000	0.073000	0.071000	0.086000	0.000001	33,34,35,36,37,38,39,40

The maximum number of final features can be adjusted using the n_filter (default=100) parameter. The actual number of features may be less, depending on: (a) the initial number of features generated (defined by the part-split-scale combinations), and (b) the strictness of both pre-filtering and filtering criteria.

# Create ~700 feature and filter them down to 10 features
df_feat = cpp.run(labels=labels, n_filter=10)
aa.display_df(df_feat, n_rows=10, show_shape=True)

DataFrame shape: (10, 13)

	feature	category	subcategory	scale_name	scale_description	abs_auc	abs_mean_dif	mean_dif	std_test	std_ref	p_val_mann_whitney	p_val_fdr_bh	positions
1	TMD_JMD-Segment...4,5)-ROBB760113	Conformation	β-turn	β-turn	Information mea...n-Suzuki, 1976)	0.316000	0.137000	-0.137000	0.102000	0.108000	0.000000	0.000000	25,26,27,28,29,30,31,32
2	TMD_JMD-Segment...4,4)-ZIMJ680104	Energy	Isoelectric point	Isoelectric point	Isoelectric poi...n et al., 1968)	0.312000	0.099000	0.099000	0.069000	0.095000	0.000000	0.000000	31,32,33,34,35,36,37,38,39,40
3	TMD_JMD-Segment...4,5)-KANM800103	Conformation	α-helix	α-helix	Average relativ...sa-Tsong, 1980)	0.297000	0.086000	0.086000	0.077000	0.068000	0.000000	0.000000	25,26,27,28,29,30,31,32
4	TMD_JMD-Segment...5,5)-LINS030104	ASA/Volume	Accessible surface area (ASA)	ASA (folded protein)	Total median ac...s et al., 2003)	0.295000	0.141000	0.141000	0.115000	0.130000	0.000000	0.000000	33,34,35,36,37,38,39,40
5	TMD_JMD-Segment...5,5)-JANJ780102	ASA/Volume	Buried	Buried	Percentage of b...n et al., 1978)	0.291000	0.130000	-0.130000	0.099000	0.124000	0.000000	0.000000	33,34,35,36,37,38,39,40
6	TMD_JMD-Segment...5,5)-ZIMJ680103	Polarity	Hydrophilicity	Polarity (hydrophilicity)	Polarity (Zimme...n et al., 1968)	0.289000	0.178000	0.178000	0.159000	0.163000	0.000000	0.000000	33,34,35,36,37,38,39,40
7	TMD_JMD-Segment...4,5)-FUKS010106	Composition	Membrane proteins (MPs)	Proteins of mesophiles (INT)	Interior compos...ishikawa, 2001)	0.277000	0.123000	0.123000	0.104000	0.127000	0.000000	0.000000	25,26,27,28,29,30,31,32
8	TMD_JMD-Segment...4,4)-WOLR790101	Polarity	Hydrophobicity (surrounding)	Hydration potential	Hydrophobicity ...n et al., 1979)	0.267000	0.105000	-0.105000	0.100000	0.113000	0.000000	0.000000	31,32,33,34,35,36,37,38,39,40
9	TMD_JMD-Segment...5,5)-MIYS990104	Composition	MPs (anchor)	Partition energy	Optimized relat...Jernigan, 1999)	0.243000	0.103000	0.103000	0.095000	0.126000	0.000002	0.000004	33,34,35,36,37,38,39,40
10	TMD_JMD-Segment...4,5)-ANDN920101	Structure-Activity	Backbone-dynamics (-CH)	α-CH chemical s...kbone-dynamics)	alpha-CH chemic...n et al., 1992)	0.229000	0.102000	-0.102000	0.097000	0.125000	0.000009	0.000012	25,26,27,28,29,30,31,32

In the initial CPP pre-filtering step, you can either set the number of retained features using n_pre_filter or define a percentage of initial features with pct_pre_filter (default with 5%). Additionally, adjust the maximum standard deviation allowed in the test dataset for each feature via max_std_test:

# Pre-filtering by allowing 50% with 0.5 maximum std in the test set
# Create ~700 feature and filter them down to 26 features
df_feat = cpp.run(labels=labels, pct_pre_filter=50, max_std_test=0.5)
aa.display_df(df_feat, n_rows=10, show_shape=True)

DataFrame shape: (26, 13)

	feature	category	subcategory	scale_name	scale_description	abs_auc	abs_mean_dif	mean_dif	std_test	std_ref	p_val_fdr_bh	positions
1	TMD_JMD-Segment...4,5)-ROBB760113	Conformation	β-turn	β-turn	Information mea...n-Suzuki, 1976)	0.316000	0.137000	-0.137000	0.102000	0.108000	0.000000	25,26,27,28,29,30,31,32
2	TMD_JMD-Segment...4,4)-ZIMJ680104	Energy	Isoelectric point	Isoelectric point	Isoelectric poi...n et al., 1968)	0.312000	0.099000	0.099000	0.069000	0.095000	0.000000	31,32,33,34,35,36,37,38,39,40
3	TMD_JMD-Segment...2,2)-ONEK900101	Others	Unclassified (Others)	ΔG values in peptides	Delta G values ...-DeGrado, 1990)	0.310000	0.041000	0.041000	0.028000	0.044000	0.000000	21,22,23,24,25,...,36,37,38,39,40
4	TMD_JMD-Segment...4,5)-KANM800103	Conformation	α-helix	α-helix	Average relativ...sa-Tsong, 1980)	0.297000	0.086000	0.086000	0.077000	0.068000	0.000000	25,26,27,28,29,30,31,32
5	TMD_JMD-Segment...5,5)-LINS030104	ASA/Volume	Accessible surface area (ASA)	ASA (folded protein)	Total median ac...s et al., 2003)	0.295000	0.141000	0.141000	0.115000	0.130000	0.000001	33,34,35,36,37,38,39,40
6	TMD_JMD-Segment...5,5)-JANJ780102	ASA/Volume	Buried	Buried	Percentage of b...n et al., 1978)	0.291000	0.130000	-0.130000	0.099000	0.124000	0.000001	33,34,35,36,37,38,39,40
7	TMD_JMD-Segment...5,5)-ZIMJ680103	Polarity	Hydrophilicity	Polarity (hydrophilicity)	Polarity (Zimme...n et al., 1968)	0.289000	0.178000	0.178000	0.159000	0.163000	0.000001	33,34,35,36,37,38,39,40
8	TMD_JMD-Segment...4,5)-FUKS010106	Composition	Membrane proteins (MPs)	Proteins of mesophiles (INT)	Interior compos...ishikawa, 2001)	0.277000	0.123000	0.123000	0.104000	0.127000	0.000001	25,26,27,28,29,30,31,32
9	TMD_JMD-Segment...3,4)-WOLR790101	Polarity	Hydrophobicity (surrounding)	Hydration potential	Hydrophobicity ...n et al., 1979)	0.274000	0.052000	0.052000	0.034000	0.060000	0.000001	21,22,23,24,25,26,27,28,29,30
10	TMD_JMD-Segment...1,2)-WEBA780101	Others	Mutability	RF value	RF value in hig...er-Lacey, 1978)	0.268000	0.042000	0.042000	0.039000	0.046000	0.000002	1,2,3,4,5,6,7,8...,16,17,18,19,20

For the final CPP filtering step, you can use the following parameters: max_overlap setting the allowed maximum positional overlap of similar features (the higher, the less strict), max_cor defining the allowed maximum Pearson correlation for scales of similar features (the higher, the less strict), check_cat setting whether redundancy of scale categories should be considered or not (setting it to False will result in stricter filtering since features across all categories are compared), and redundancy selecting the position-overlap criterion ('legacy' by default, kept so results stay reproducible; 'exact' compares the actual residue positions):

# Disable filtering by setting max_overlap and max_cor to 1
# Create ~700 feature and filter them down to 100 features
df_feat = cpp.run(labels=labels, max_overlap=1, max_cor=1)
aa.display_df(df_feat, n_rows=10, show_shape=True)

DataFrame shape: (100, 13)

	feature	category	subcategory	scale_name	scale_description	abs_auc	abs_mean_dif	mean_dif	std_test	std_ref	positions
1	TMD_JMD-Segment...4,5)-ROBB760113	Conformation	β-turn	β-turn	Information mea...n-Suzuki, 1976)	0.316000	0.137000	-0.137000	0.102000	0.108000	25,26,27,28,29,30,31,32
2	TMD_JMD-Segment...4,4)-ZIMJ680104	Energy	Isoelectric point	Isoelectric point	Isoelectric poi...n et al., 1968)	0.312000	0.099000	0.099000	0.069000	0.095000	31,32,33,34,35,36,37,38,39,40
3	TMD_JMD-Segment...3,3)-ZIMJ680104	Energy	Isoelectric point	Isoelectric point	Isoelectric poi...n et al., 1968)	0.304000	0.069000	0.069000	0.051000	0.073000	27,28,29,30,31,...,36,37,38,39,40
4	TMD_JMD-Segment...4,5)-KANM800103	Conformation	α-helix	α-helix	Average relativ...sa-Tsong, 1980)	0.297000	0.086000	0.086000	0.077000	0.068000	25,26,27,28,29,30,31,32
5	TMD_JMD-Segment...5,5)-LINS030104	ASA/Volume	Accessible surface area (ASA)	ASA (folded protein)	Total median ac...s et al., 2003)	0.295000	0.141000	0.141000	0.115000	0.130000	33,34,35,36,37,38,39,40
6	TMD_JMD-Segment...2,2)-KANM800103	Conformation	α-helix	α-helix	Average relativ...sa-Tsong, 1980)	0.292000	0.058000	0.058000	0.045000	0.054000	21,22,23,24,25,...,36,37,38,39,40
7	TMD_JMD-Segment...5,5)-JANJ780102	ASA/Volume	Buried	Buried	Percentage of b...n et al., 1978)	0.291000	0.130000	-0.130000	0.099000	0.124000	33,34,35,36,37,38,39,40
8	TMD_JMD-Segment...4,4)-LINS030104	ASA/Volume	Accessible surface area (ASA)	ASA (folded protein)	Total median ac...s et al., 2003)	0.291000	0.127000	0.127000	0.097000	0.121000	31,32,33,34,35,36,37,38,39,40
9	TMD_JMD-Segment...5,5)-ZIMJ680103	Polarity	Hydrophilicity	Polarity (hydrophilicity)	Polarity (Zimme...n et al., 1968)	0.289000	0.178000	0.178000	0.159000	0.163000	33,34,35,36,37,38,39,40
10	TMD_JMD-Segment...4,4)-ZIMJ680103	Polarity	Hydrophilicity	Polarity (hydrophilicity)	Polarity (Zimme...n et al., 1968)	0.288000	0.164000	0.164000	0.135000	0.145000	31,32,33,34,35,36,37,38,39,40

# Perform stricter filtering by setting check_cat=False
# Create ~700 feature and filter them down to 11 features
df_feat = cpp.run(labels=labels, check_cat=False)
aa.display_df(df_feat, n_rows=10, show_shape=True)

DataFrame shape: (11, 13)

	feature	category	subcategory	scale_name	scale_description	abs_auc	abs_mean_dif	mean_dif	std_test	std_ref	p_val_mann_whitney	p_val_fdr_bh	positions
1	TMD_JMD-Segment...4,5)-ROBB760113	Conformation	β-turn	β-turn	Information mea...n-Suzuki, 1976)	0.316000	0.137000	-0.137000	0.102000	0.108000	0.000000	0.000000	25,26,27,28,29,30,31,32
2	TMD_JMD-Segment...4,4)-ZIMJ680104	Energy	Isoelectric point	Isoelectric point	Isoelectric poi...n et al., 1968)	0.312000	0.099000	0.099000	0.069000	0.095000	0.000000	0.000000	31,32,33,34,35,36,37,38,39,40
3	TMD_JMD-Segment...4,5)-KANM800103	Conformation	α-helix	α-helix	Average relativ...sa-Tsong, 1980)	0.297000	0.086000	0.086000	0.077000	0.068000	0.000000	0.000000	25,26,27,28,29,30,31,32
4	TMD_JMD-Segment...5,5)-LINS030104	ASA/Volume	Accessible surface area (ASA)	ASA (folded protein)	Total median ac...s et al., 2003)	0.295000	0.141000	0.141000	0.115000	0.130000	0.000000	0.000000	33,34,35,36,37,38,39,40
5	TMD_JMD-Segment...5,5)-JANJ780102	ASA/Volume	Buried	Buried	Percentage of b...n et al., 1978)	0.291000	0.130000	-0.130000	0.099000	0.124000	0.000000	0.000000	33,34,35,36,37,38,39,40
6	TMD_JMD-Segment...2,2)-CEDJ970105	Composition	AA composition	Nuclear proteins	Composition of ...o et al., 1997)	0.263000	0.062000	0.062000	0.062000	0.069000	0.000000	0.000001	21,22,23,24,25,...,36,37,38,39,40
7	TMD_JMD-Segment...5,5)-MITS020101	Polarity	Amphiphilicity	Amphiphilicity	Amphiphilicity ...u et al., 2002)	0.262000	0.073000	0.073000	0.071000	0.086000	0.000000	0.000001	33,34,35,36,37,38,39,40
8	TMD_JMD-Segment...1,2)-SIMZ760101	Polarity	Hydrophobicity	Transfer free e...TFE) to outside	Transfer free e...-Charton (1982)	0.259000	0.064000	-0.064000	0.069000	0.072000	0.000001	0.000002	1,2,3,4,5,6,7,8...,16,17,18,19,20
9	TMD_JMD-Segment...4,5)-ANDN920101	Structure-Activity	Backbone-dynamics (-CH)	α-CH chemical s...kbone-dynamics)	alpha-CH chemic...n et al., 1992)	0.229000	0.102000	-0.102000	0.097000	0.125000	0.000009	0.000017	25,26,27,28,29,30,31,32
10	TMD_JMD-Segment...4,4)-YUTK870103	Energy	Free energy (unfolding)	Free energy (unfolding)	Activation Gibb...i et al., 1987)	0.201000	0.084000	-0.084000	0.115000	0.118000	0.000103	0.000143	31,32,33,34,35,36,37,38,39,40

Switch the position-overlap criterion to 'exact'. This is an enhancement, not a bug fix: it compares the true residue positions and tends to yield a more concentrated signature (fewer redundant subcategories), while predictive performance is essentially unchanged. For a stronger, more efficient redundancy reduction, see CPP().simplify():

# Switch the position-overlap criterion to the exact residue positions
# (default is "legacy", kept for reproducibility)
df_feat = cpp.run(labels=labels, redundancy="exact")
aa.display_df(df_feat, n_rows=10, show_shape=True)

DataFrame shape: (29, 13)

	feature	category	subcategory	scale_name	scale_description	abs_auc	abs_mean_dif	mean_dif	std_test	std_ref	p_val_fdr_bh	positions
1	TMD_JMD-Segment...4,5)-ROBB760113	Conformation	β-turn	β-turn	Information mea...n-Suzuki, 1976)	0.316000	0.137000	-0.137000	0.102000	0.108000	0.000000	25,26,27,28,29,30,31,32
2	TMD_JMD-Segment...4,4)-ZIMJ680104	Energy	Isoelectric point	Isoelectric point	Isoelectric poi...n et al., 1968)	0.312000	0.099000	0.099000	0.069000	0.095000	0.000000	31,32,33,34,35,36,37,38,39,40
3	TMD_JMD-Segment...4,5)-KANM800103	Conformation	α-helix	α-helix	Average relativ...sa-Tsong, 1980)	0.297000	0.086000	0.086000	0.077000	0.068000	0.000000	25,26,27,28,29,30,31,32
4	TMD_JMD-Segment...5,5)-LINS030104	ASA/Volume	Accessible surface area (ASA)	ASA (folded protein)	Total median ac...s et al., 2003)	0.295000	0.141000	0.141000	0.115000	0.130000	0.000000	33,34,35,36,37,38,39,40
5	TMD_JMD-Segment...2,2)-KANM800103	Conformation	α-helix	α-helix	Average relativ...sa-Tsong, 1980)	0.292000	0.058000	0.058000	0.045000	0.054000	0.000000	21,22,23,24,25,...,36,37,38,39,40
6	TMD_JMD-Segment...5,5)-JANJ780102	ASA/Volume	Buried	Buried	Percentage of b...n et al., 1978)	0.291000	0.130000	-0.130000	0.099000	0.124000	0.000000	33,34,35,36,37,38,39,40
7	TMD_JMD-Segment...5,5)-ZIMJ680103	Polarity	Hydrophilicity	Polarity (hydrophilicity)	Polarity (Zimme...n et al., 1968)	0.289000	0.178000	0.178000	0.159000	0.163000	0.000000	33,34,35,36,37,38,39,40
8	TMD_JMD-Segment...2,2)-CHOP780212	Conformation	β-sheet (C-term)	β-turn (1st residue)	Frequency of th...-Fasman, 1978b)	0.280000	0.066000	-0.066000	0.048000	0.067000	0.000000	21,22,23,24,25,...,36,37,38,39,40
9	TMD_JMD-Segment...4,5)-FUKS010106	Composition	Membrane proteins (MPs)	Proteins of mesophiles (INT)	Interior compos...ishikawa, 2001)	0.277000	0.123000	0.123000	0.104000	0.127000	0.000000	25,26,27,28,29,30,31,32
10	TMD_JMD-Segment...4,4)-WOLR790101	Polarity	Hydrophobicity (surrounding)	Hydration potential	Hydrophobicity ...n et al., 1979)	0.267000	0.105000	-0.105000	0.100000	0.113000	0.000001	31,32,33,34,35,36,37,38,39,40

The residue positions can be adjusted using the start, tmd_len, jmd_n_len, and jmd_c_len parameters:

# Shift positions by 10 residues
df_feat = cpp.run(labels=labels, start=11)
aa.display_df(df_feat, n_rows=10, show_shape=True)

DataFrame shape: (19, 13)

	feature	category	subcategory	scale_name	scale_description	abs_auc	abs_mean_dif	mean_dif	std_test	std_ref	p_val_fdr_bh	positions
1	TMD_JMD-Segment...4,5)-ROBB760113	Conformation	β-turn	β-turn	Information mea...n-Suzuki, 1976)	0.316000	0.137000	-0.137000	0.102000	0.108000	0.000000	35,36,37,38,39,40,41,42
2	TMD_JMD-Segment...4,4)-ZIMJ680104	Energy	Isoelectric point	Isoelectric point	Isoelectric poi...n et al., 1968)	0.312000	0.099000	0.099000	0.069000	0.095000	0.000000	41,42,43,44,45,46,47,48,49,50
3	TMD_JMD-Segment...4,5)-KANM800103	Conformation	α-helix	α-helix	Average relativ...sa-Tsong, 1980)	0.297000	0.086000	0.086000	0.077000	0.068000	0.000000	35,36,37,38,39,40,41,42
4	TMD_JMD-Segment...5,5)-LINS030104	ASA/Volume	Accessible surface area (ASA)	ASA (folded protein)	Total median ac...s et al., 2003)	0.295000	0.141000	0.141000	0.115000	0.130000	0.000000	43,44,45,46,47,48,49,50
5	TMD_JMD-Segment...5,5)-JANJ780102	ASA/Volume	Buried	Buried	Percentage of b...n et al., 1978)	0.291000	0.130000	-0.130000	0.099000	0.124000	0.000000	43,44,45,46,47,48,49,50
6	TMD_JMD-Segment...5,5)-ZIMJ680103	Polarity	Hydrophilicity	Polarity (hydrophilicity)	Polarity (Zimme...n et al., 1968)	0.289000	0.178000	0.178000	0.159000	0.163000	0.000000	43,44,45,46,47,48,49,50
7	TMD_JMD-Segment...4,5)-FUKS010106	Composition	Membrane proteins (MPs)	Proteins of mesophiles (INT)	Interior compos...ishikawa, 2001)	0.277000	0.123000	0.123000	0.104000	0.127000	0.000000	35,36,37,38,39,40,41,42
8	TMD_JMD-Segment...4,4)-WOLR790101	Polarity	Hydrophobicity (surrounding)	Hydration potential	Hydrophobicity ...n et al., 1979)	0.267000	0.105000	-0.105000	0.100000	0.113000	0.000001	41,42,43,44,45,46,47,48,49,50
9	TMD_JMD-Segment...2,2)-CEDJ970105	Composition	AA composition	Nuclear proteins	Composition of ...o et al., 1997)	0.263000	0.062000	0.062000	0.062000	0.069000	0.000001	31,32,33,34,35,...,46,47,48,49,50
10	TMD_JMD-Segment...5,5)-MITS020101	Polarity	Amphiphilicity	Amphiphilicity	Amphiphilicity ...u et al., 2002)	0.262000	0.073000	0.073000	0.071000	0.086000	0.000001	43,44,45,46,47,48,49,50

# Increase TMD length from 20 to 50
df_feat = cpp.run(labels=labels, tmd_len=50)
aa.display_df(df_feat, n_rows=10, show_shape=True)

DataFrame shape: (19, 13)

	feature	category	subcategory	scale_name	scale_description	abs_auc	abs_mean_dif	mean_dif	std_test	std_ref	p_val_fdr_bh	positions
1	TMD_JMD-Segment...4,5)-ROBB760113	Conformation	β-turn	β-turn	Information mea...n-Suzuki, 1976)	0.316000	0.137000	-0.137000	0.102000	0.108000	0.000000	43,44,45,46,47,...,52,53,54,55,56
2	TMD_JMD-Segment...4,4)-ZIMJ680104	Energy	Isoelectric point	Isoelectric point	Isoelectric poi...n et al., 1968)	0.312000	0.099000	0.099000	0.069000	0.095000	0.000000	53,54,55,56,57,...,66,67,68,69,70
3	TMD_JMD-Segment...4,5)-KANM800103	Conformation	α-helix	α-helix	Average relativ...sa-Tsong, 1980)	0.297000	0.086000	0.086000	0.077000	0.068000	0.000000	43,44,45,46,47,...,52,53,54,55,56
4	TMD_JMD-Segment...5,5)-LINS030104	ASA/Volume	Accessible surface area (ASA)	ASA (folded protein)	Total median ac...s et al., 2003)	0.295000	0.141000	0.141000	0.115000	0.130000	0.000000	57,58,59,60,61,...,66,67,68,69,70
5	TMD_JMD-Segment...5,5)-JANJ780102	ASA/Volume	Buried	Buried	Percentage of b...n et al., 1978)	0.291000	0.130000	-0.130000	0.099000	0.124000	0.000000	57,58,59,60,61,...,66,67,68,69,70
6	TMD_JMD-Segment...5,5)-ZIMJ680103	Polarity	Hydrophilicity	Polarity (hydrophilicity)	Polarity (Zimme...n et al., 1968)	0.289000	0.178000	0.178000	0.159000	0.163000	0.000000	57,58,59,60,61,...,66,67,68,69,70
7	TMD_JMD-Segment...4,5)-FUKS010106	Composition	Membrane proteins (MPs)	Proteins of mesophiles (INT)	Interior compos...ishikawa, 2001)	0.277000	0.123000	0.123000	0.104000	0.127000	0.000000	43,44,45,46,47,...,52,53,54,55,56
8	TMD_JMD-Segment...4,4)-WOLR790101	Polarity	Hydrophobicity (surrounding)	Hydration potential	Hydrophobicity ...n et al., 1979)	0.267000	0.105000	-0.105000	0.100000	0.113000	0.000001	53,54,55,56,57,...,66,67,68,69,70
9	TMD_JMD-Segment...2,2)-CEDJ970105	Composition	AA composition	Nuclear proteins	Composition of ...o et al., 1997)	0.263000	0.062000	0.062000	0.062000	0.069000	0.000001	36,37,38,39,40,...,66,67,68,69,70
10	TMD_JMD-Segment...5,5)-MITS020101	Polarity	Amphiphilicity	Amphiphilicity	Amphiphilicity ...u et al., 2002)	0.262000	0.073000	0.073000	0.071000	0.086000	0.000001	57,58,59,60,61,...,66,67,68,69,70

Multiprocessing can be enabled by using the n_jobs parameter, which is set to the maximum if n_jobs=None. However, this is only recommend for more than ~1000 features per core due to potential process management overhead.

import time

# Run without multiprocessing
time_start = time.time()
df_feat = cpp.run(labels=labels, n_jobs=1)
time_no_mp = round(time.time() - time_start, 2)
print(f"Time without multiprocessing: {time_no_mp} seconds")

# Run with multiprocessing
time_start = time.time()
df_feat = cpp.run(labels=labels, n_jobs=None)
time_mp = round(time.time() - time_start, 2)
print(f"Time with multiprocessing. {time_mp} seconds")

Time without multiprocessing: 0.03 seconds
Time with multiprocessing. 3.93 seconds

run exposes several more options. label_test / label_ref name the two groups in labels (default 1 / 0); parametric switches the p-value test (independent t-test instead of the Mann-Whitney U test); return_stats also returns the filter-funnel statistics; and vectorized, n_batches and n_sample_batches trade memory for speed on large inputs:

df_feat, stats = cpp.run(labels=labels, label_test=1, label_ref=0,
                         parametric=True, vectorized=True, return_stats=True)
print("filter stats:", stats)
aa.display_df(df_feat, n_rows=5, show_shape=True)

filter stats: {'n_candidates': 690, 'n_after_prefilter': 100, 'n_after_redundancy': 19, 'n_final': 19}
DataFrame shape: (19, 13)

	feature	category	subcategory	scale_name	scale_description	abs_auc	abs_mean_dif	mean_dif	std_test	std_ref	positions
1	TMD_JMD-Segment...4,5)-ROBB760113	Conformation	β-turn	β-turn	Information mea...n-Suzuki, 1976)	0.316000	0.137000	-0.137000	0.102000	0.108000	25,26,27,28,29,30,31,32
2	TMD_JMD-Segment...4,4)-ZIMJ680104	Energy	Isoelectric point	Isoelectric point	Isoelectric poi...n et al., 1968)	0.312000	0.099000	0.099000	0.069000	0.095000	31,32,33,34,35,36,37,38,39,40
3	TMD_JMD-Segment...4,5)-KANM800103	Conformation	α-helix	α-helix	Average relativ...sa-Tsong, 1980)	0.297000	0.086000	0.086000	0.077000	0.068000	25,26,27,28,29,30,31,32
4	TMD_JMD-Segment...5,5)-LINS030104	ASA/Volume	Accessible surface area (ASA)	ASA (folded protein)	Total median ac...s et al., 2003)	0.295000	0.141000	0.141000	0.115000	0.130000	33,34,35,36,37,38,39,40
5	TMD_JMD-Segment...5,5)-JANJ780102	ASA/Volume	Buried	Buried	Percentage of b...n et al., 1978)	0.291000	0.130000	-0.130000	0.099000	0.124000	33,34,35,36,37,38,39,40

Further parameters. run also exposes the JMD geometry (jmd_n_len / jmd_c_len), an explicit pre-filter count n_pre_filter (instead of the pct_pre_filter percentage), and two mutually exclusive memory-bounding batch axes: n_batches (over scales) and n_sample_batches (over samples):

# JMD geometry + an explicit pre-filter count + scale-axis batching (n_batches, >=2)
df_feat = cpp.run(labels=labels, jmd_n_len=10, jmd_c_len=10,
                  n_pre_filter=100, n_batches=2, n_jobs=1)
aa.display_df(df_feat, n_rows=10, show_shape=True)

# Sample-axis batching bounds peak memory by batch size (n_sample_batches, >=2);
# mutually exclusive with n_batches
df_feat = cpp.run(labels=labels, n_sample_batches=2, n_jobs=1)
aa.display_df(df_feat, n_rows=10, show_shape=True)

DataFrame shape: (19, 13)

	feature	category	subcategory	scale_name	scale_description	abs_auc	abs_mean_dif	mean_dif	std_test	std_ref	p_val_fdr_bh	positions
1	TMD_JMD-Segment...4,5)-ROBB760113	Conformation	β-turn	β-turn	Information mea...n-Suzuki, 1976)	0.316000	0.137000	-0.137000	0.102000	0.108000	0.000000	25,26,27,28,29,30,31,32
2	TMD_JMD-Segment...4,4)-ZIMJ680104	Energy	Isoelectric point	Isoelectric point	Isoelectric poi...n et al., 1968)	0.312000	0.099000	0.099000	0.069000	0.095000	0.000000	31,32,33,34,35,36,37,38,39,40
3	TMD_JMD-Segment...4,5)-KANM800103	Conformation	α-helix	α-helix	Average relativ...sa-Tsong, 1980)	0.297000	0.086000	0.086000	0.077000	0.068000	0.000000	25,26,27,28,29,30,31,32
4	TMD_JMD-Segment...5,5)-LINS030104	ASA/Volume	Accessible surface area (ASA)	ASA (folded protein)	Total median ac...s et al., 2003)	0.295000	0.141000	0.141000	0.115000	0.130000	0.000000	33,34,35,36,37,38,39,40
5	TMD_JMD-Segment...5,5)-JANJ780102	ASA/Volume	Buried	Buried	Percentage of b...n et al., 1978)	0.291000	0.130000	-0.130000	0.099000	0.124000	0.000000	33,34,35,36,37,38,39,40
6	TMD_JMD-Segment...5,5)-ZIMJ680103	Polarity	Hydrophilicity	Polarity (hydrophilicity)	Polarity (Zimme...n et al., 1968)	0.289000	0.178000	0.178000	0.159000	0.163000	0.000000	33,34,35,36,37,38,39,40
7	TMD_JMD-Segment...4,5)-FUKS010106	Composition	Membrane proteins (MPs)	Proteins of mesophiles (INT)	Interior compos...ishikawa, 2001)	0.277000	0.123000	0.123000	0.104000	0.127000	0.000000	25,26,27,28,29,30,31,32
8	TMD_JMD-Segment...4,4)-WOLR790101	Polarity	Hydrophobicity (surrounding)	Hydration potential	Hydrophobicity ...n et al., 1979)	0.267000	0.105000	-0.105000	0.100000	0.113000	0.000001	31,32,33,34,35,36,37,38,39,40
9	TMD_JMD-Segment...2,2)-CEDJ970105	Composition	AA composition	Nuclear proteins	Composition of ...o et al., 1997)	0.263000	0.062000	0.062000	0.062000	0.069000	0.000003	21,22,23,24,25,...,36,37,38,39,40
10	TMD_JMD-Segment...5,5)-MITS020101	Polarity	Amphiphilicity	Amphiphilicity	Amphiphilicity ...u et al., 2002)	0.262000	0.073000	0.073000	0.071000	0.086000	0.000001	33,34,35,36,37,38,39,40

DataFrame shape: (19, 13)

	feature	category	subcategory	scale_name	scale_description	abs_auc	abs_mean_dif	mean_dif	std_test	std_ref	p_val_fdr_bh	positions
1	TMD_JMD-Segment...4,5)-ROBB760113	Conformation	β-turn	β-turn	Information mea...n-Suzuki, 1976)	0.316000	0.137000	-0.137000	0.102000	0.108000	0.000000	25,26,27,28,29,30,31,32
2	TMD_JMD-Segment...4,4)-ZIMJ680104	Energy	Isoelectric point	Isoelectric point	Isoelectric poi...n et al., 1968)	0.312000	0.099000	0.099000	0.069000	0.095000	0.000000	31,32,33,34,35,36,37,38,39,40
3	TMD_JMD-Segment...4,5)-KANM800103	Conformation	α-helix	α-helix	Average relativ...sa-Tsong, 1980)	0.297000	0.086000	0.086000	0.077000	0.068000	0.000000	25,26,27,28,29,30,31,32
4	TMD_JMD-Segment...5,5)-LINS030104	ASA/Volume	Accessible surface area (ASA)	ASA (folded protein)	Total median ac...s et al., 2003)	0.295000	0.141000	0.141000	0.115000	0.130000	0.000000	33,34,35,36,37,38,39,40
5	TMD_JMD-Segment...5,5)-JANJ780102	ASA/Volume	Buried	Buried	Percentage of b...n et al., 1978)	0.291000	0.130000	-0.130000	0.099000	0.124000	0.000000	33,34,35,36,37,38,39,40
6	TMD_JMD-Segment...5,5)-ZIMJ680103	Polarity	Hydrophilicity	Polarity (hydrophilicity)	Polarity (Zimme...n et al., 1968)	0.289000	0.178000	0.178000	0.159000	0.163000	0.000000	33,34,35,36,37,38,39,40
7	TMD_JMD-Segment...4,5)-FUKS010106	Composition	Membrane proteins (MPs)	Proteins of mesophiles (INT)	Interior compos...ishikawa, 2001)	0.277000	0.123000	0.123000	0.104000	0.127000	0.000000	25,26,27,28,29,30,31,32
8	TMD_JMD-Segment...4,4)-WOLR790101	Polarity	Hydrophobicity (surrounding)	Hydration potential	Hydrophobicity ...n et al., 1979)	0.267000	0.105000	-0.105000	0.100000	0.113000	0.000001	31,32,33,34,35,36,37,38,39,40
9	TMD_JMD-Segment...2,2)-CEDJ970105	Composition	AA composition	Nuclear proteins	Composition of ...o et al., 1997)	0.263000	0.062000	0.062000	0.062000	0.069000	0.000001	21,22,23,24,25,...,36,37,38,39,40
10	TMD_JMD-Segment...5,5)-MITS020101	Polarity	Amphiphilicity	Amphiphilicity	Amphiphilicity ...u et al., 2002)	0.262000	0.073000	0.073000	0.071000	0.086000	0.000001	33,34,35,36,37,38,39,40

run_num is the numerical-mode twin of run: per-residue values come from a pre-sliced tensor dict_num_parts (built by :meth:NumericalFeature.get_parts) instead of an amino-acid-to-scale lookup. Use it for protein language model embeddings, structure features, or any per-residue numerical representation:

import numpy as np
nf = aa.NumericalFeature()
# One [0, 1]-normalized per-residue tensor (here 2 embedding-like dimensions) per entry
rng = np.random.default_rng(0)
dict_num = {row.entry: rng.random((len(row.sequence), 2)) for row in df_seq.itertuples()}
df_parts_num, dict_num_parts = nf.get_parts(df_seq=df_seq, dict_num=dict_num)
# df_scales columns name the D dimensions of dict_num_parts (2 here)
df_scales_num = aa.load_scales().iloc[:, :2]
cpp = aa.CPP(df_parts=df_parts_num, df_scales=df_scales_num)
df_feat_num = cpp.run_num(dict_num_parts=dict_num_parts, labels=labels,
                              n_filter=10, n_jobs=1)
aa.display_df(df_feat_num, n_rows=10, show_shape=True)

DataFrame shape: (10, 13)

	feature	category	subcategory	scale_name	scale_description	abs_auc	abs_mean_dif	mean_dif	std_test	std_ref	p_val_mann_whitney	p_val_fdr_bh	positions
1	TMD_C_JMD_C-Pat...,12)-ANDN920101	Structure-Activity	Backbone-dynamics (-CH)	α-CH chemical s...kbone-dynamics)	alpha-CH chemic...n et al., 1992)	0.201000	0.108000	0.108000	0.129000	0.149000	0.000096	0.009487	29,32,36,40
2	TMD_C_JMD_C-Pat...,13)-ANDN920101	Structure-Activity	Backbone-dynamics (-CH)	α-CH chemical s...kbone-dynamics)	alpha-CH chemic...n et al., 1992)	0.150000	0.069000	0.069000	0.139000	0.134000	0.003666	0.035587	28,32,35,38
3	TMD_C_JMD_C-Pat...,13)-ARGP820101	Polarity	Hydrophobicity	Hydrophobicity	Hydrophobicity ...s et al., 1982)	0.137000	0.064000	0.064000	0.128000	0.129000	0.007890	0.038845	28,32,36,40
4	TMD-Pattern(C,1...,14)-ARGP820101	Polarity	Hydrophobicity	Hydrophobicity	Hydrophobicity ...s et al., 1982)	0.124000	0.091000	-0.091000	0.196000	0.209000	0.016261	0.047435	17,21
5	TMD_C_JMD_C-Pat...,10)-ANDN920101	Structure-Activity	Backbone-dynamics (-CH)	α-CH chemical s...kbone-dynamics)	alpha-CH chemic...n et al., 1992)	0.122000	0.064000	-0.064000	0.159000	0.153000	0.018320	0.047435	31,34,37
6	TMD_C_JMD_C-Pat...,14)-ARGP820101	Polarity	Hydrophobicity	Hydrophobicity	Hydrophobicity ...s et al., 1982)	0.111000	0.084000	-0.084000	0.200000	0.205000	0.031612	0.059048	31,34
7	TMD-Segment(8,9)-ANDN920101	Structure-Activity	Backbone-dynamics (-CH)	α-CH chemical s...kbone-dynamics)	alpha-CH chemic...n et al., 1992)	0.110000	0.057000	0.057000	0.147000	0.174000	0.033596	0.060472	26,27
8	JMD_N_TMD_N-Pat...,15)-ANDN920101	Structure-Activity	Backbone-dynamics (-CH)	α-CH chemical s...kbone-dynamics)	alpha-CH chemic...n et al., 1992)	0.103000	0.068000	0.068000	0.189000	0.184000	0.045713	0.073752	6,10
9	JMD_N_TMD_N-Pat...,12)-ANDN920101	Structure-Activity	Backbone-dynamics (-CH)	α-CH chemical s...kbone-dynamics)	alpha-CH chemic...n et al., 1992)	0.099000	0.057000	0.057000	0.168000	0.149000	0.054555	0.077156	4,8,12
10	TMD_C_JMD_C-Pat...,13)-ARGP820101	Polarity	Hydrophobicity	Hydrophobicity	Hydrophobicity ...s et al., 1982)	0.096000	0.056000	0.056000	0.174000	0.206000	0.063378	0.082383	29,33