Spaces:
Running
Running
File size: 12,290 Bytes
f4b267d c95d941 f4b267d c95d941 f4b267d | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 | """
Tests for protein_conformal/util.py core functions.
These tests verify:
1. FASTA parsing
2. FAISS database operations
3. FDR/FNR threshold calculations (conformal risk control)
4. Risk metrics (FDR, FNR, TPR)
5. Venn-Abers probability predictions
6. Hierarchical loss functions (for SCOPe)
"""
import numpy as np
import pytest
from protein_conformal.util import (
read_fasta,
load_database,
query,
get_thresh_new,
get_thresh_new_FDR,
get_thresh_FDR,
risk,
risk_1d,
calculate_false_negatives,
calculate_true_positives,
simplifed_venn_abers_prediction,
get_isotone_regression,
scope_hierarchical_loss,
validate_lhat_new,
)
class TestFastaParsing:
"""Tests for FASTA file parsing."""
def test_read_fasta_basic(self, sample_fasta_file):
"""Test basic FASTA parsing returns sequences and metadata."""
sequences, metadata = read_fasta(sample_fasta_file)
assert len(sequences) == 3
assert len(metadata) == 3
# Check first sequence
assert sequences[0].startswith('MVLSPADKTN')
assert '>protein1' in metadata[0]
def test_read_fasta_sequence_content(self, sample_fasta_file):
"""Test that sequences contain only valid amino acids."""
sequences, _ = read_fasta(sample_fasta_file)
valid_aa = set('ACDEFGHIKLMNPQRSTVWY')
for seq in sequences:
assert all(aa in valid_aa for aa in seq), f"Invalid AA in sequence: {seq}"
def test_read_fasta_short_sequence(self, sample_fasta_file):
"""Test that short sequence is parsed correctly."""
sequences, metadata = read_fasta(sample_fasta_file)
# Third sequence is exactly the 20 standard amino acids
assert sequences[2] == 'ACDEFGHIKLMNPQRSTVWY'
assert len(sequences[2]) == 20
class TestFAISSOperations:
"""Tests for FAISS database loading and querying."""
def test_load_database(self, sample_embeddings):
"""Test that database loads and has correct dimensions."""
_, lookup_embeddings = sample_embeddings
index = load_database(lookup_embeddings.copy())
assert index.ntotal == 100 # Number of vectors in index
assert index.d == 128 # Dimensionality
def test_query_returns_correct_shape(self, sample_embeddings):
"""Test that query returns distances and indices with correct shapes."""
query_embeddings, lookup_embeddings = sample_embeddings
index = load_database(lookup_embeddings.copy())
D, I = query(index, query_embeddings.copy(), k=10)
assert D.shape == (10, 10) # 10 queries, k=10 neighbors
assert I.shape == (10, 10)
def test_query_distances_are_similarities(self, sample_embeddings):
"""Test that distances are cosine similarities (normalized dot product)."""
query_embeddings, lookup_embeddings = sample_embeddings
index = load_database(lookup_embeddings.copy())
D, I = query(index, query_embeddings.copy(), k=10)
# Cosine similarities should be in [-1, 1] range
assert D.min() >= -1.0
assert D.max() <= 1.0
def test_query_indices_valid(self, sample_embeddings):
"""Test that returned indices are valid."""
query_embeddings, lookup_embeddings = sample_embeddings
index = load_database(lookup_embeddings.copy())
D, I = query(index, query_embeddings.copy(), k=10)
# All indices should be in valid range
assert I.min() >= 0
assert I.max() < 100 # lookup has 100 embeddings
class TestRiskMetrics:
"""Tests for FDR, FNR, and related risk calculations."""
def test_risk_all_correct(self):
"""Test risk is 0 when all predictions above threshold are correct."""
sims = np.array([[0.9, 0.8, 0.7, 0.6]])
labels = np.array([[True, True, True, False]]) # First 3 are true matches
# Threshold 0.65: returns indices 0,1,2 (all true) → FDR = 0
fdr = risk(sims, labels, 0.65)
assert fdr == 0.0
def test_risk_all_incorrect(self):
"""Test risk is 1 when all predictions above threshold are incorrect."""
sims = np.array([[0.9, 0.8, 0.7, 0.6]])
labels = np.array([[False, False, False, True]]) # Only index 3 is true
# Threshold 0.65: returns indices 0,1,2 (all false) → FDR = 1
fdr = risk(sims, labels, 0.65)
assert fdr == 1.0
def test_risk_partial(self):
"""Test risk calculation with mixed predictions."""
sims = np.array([[0.9, 0.8, 0.7, 0.6]])
labels = np.array([[True, False, True, False]])
# Threshold 0.65: returns 3 items, 1 false → FDR = 1/3
fdr = risk(sims, labels, 0.65)
assert abs(fdr - 1/3) < 1e-6
def test_calculate_false_negatives_zero(self):
"""Test FNR is 0 when all positives are detected."""
sims = np.array([[0.9, 0.8, 0.7, 0.6]])
labels = np.array([[True, True, False, False]])
# Threshold 0.75: detects both true positives → FNR = 0
fnr = calculate_false_negatives(sims, labels, 0.75)
assert fnr == 0.0
def test_calculate_false_negatives_partial(self):
"""Test FNR when some positives are missed."""
sims = np.array([[0.9, 0.8, 0.7, 0.6]])
labels = np.array([[True, True, True, False]])
# Threshold 0.85: only detects index 0, misses 1,2 → FNR = 2/3
fnr = calculate_false_negatives(sims, labels, 0.85)
assert abs(fnr - 2/3) < 1e-6
class TestConformalThresholds:
"""Tests for conformal risk control threshold calculations."""
def test_get_thresh_new_basic(self, scope_like_data):
"""Test basic threshold calculation for FNR control."""
sims, labels = scope_like_data
alpha = 0.1
lhat = get_thresh_new(sims, labels, alpha)
# Threshold should be in valid similarity range
assert sims.min() <= lhat <= sims.max()
def test_get_thresh_new_FDR_basic(self, scope_like_data):
"""Test basic threshold calculation for FDR control."""
sims, labels = scope_like_data
alpha = 0.1
lhat = get_thresh_new_FDR(sims, labels, alpha)
# Threshold should be in valid similarity range
assert sims.min() <= lhat <= sims.max()
def test_threshold_decreases_with_lower_alpha(self, scope_like_data):
"""Test that more stringent alpha leads to lower threshold for FNR control.
For FNR (false negative rate) control via get_thresh_new:
- Lower alpha = more stringent = want fewer false negatives
- Algorithm picks a lower quantile of positive similarities
- Lower quantile = lower threshold = accept more matches
"""
sims, labels = scope_like_data
lhat_10 = get_thresh_new(sims, labels, alpha=0.1)
lhat_05 = get_thresh_new(sims, labels, alpha=0.05)
# Lower alpha (more stringent FNR) should give lower threshold
assert lhat_05 <= lhat_10
def test_get_thresh_FDR_returns_risk(self, scope_like_data):
"""Test that get_thresh_FDR returns both threshold and risk."""
sims, labels = scope_like_data
alpha = 0.1
lhat, risk_fdr = get_thresh_FDR(labels, sims, alpha, delta=0.5, N=100)
# Should return valid threshold and risk
assert isinstance(lhat, (int, float))
assert isinstance(risk_fdr, (int, float))
assert 0 <= risk_fdr <= 1
class TestVennAbers:
"""Tests for Venn-Abers probability predictions."""
def test_simplified_venn_abers_returns_two_probs(self):
"""Test that simplified Venn-Abers returns p0 and p1."""
np.random.seed(42)
X_cal = np.random.uniform(0.5, 1.0, 100)
Y_cal = (X_cal > 0.7).astype(bool)
X_test = 0.8
p0, p1 = simplifed_venn_abers_prediction(X_cal, Y_cal, X_test)
assert 0 <= p0 <= 1
assert 0 <= p1 <= 1
def test_venn_abers_high_similarity_high_prob(self):
"""Test that high similarity gives high probability."""
# Calibration: high sim → positive label
X_cal = np.array([0.5, 0.6, 0.7, 0.8, 0.9, 0.95])
Y_cal = np.array([False, False, False, True, True, True])
# Test point with high similarity should get high probability
p0, p1 = simplifed_venn_abers_prediction(X_cal.copy(), Y_cal.copy(), 0.92)
# Average of p0, p1 should be high for high similarity
avg_prob = (p0 + p1) / 2
assert avg_prob > 0.5
def test_isotonic_regression_monotonic(self):
"""Test that isotonic regression produces monotonic predictions."""
X = np.array([0.5, 0.6, 0.7, 0.8, 0.9])
y = np.array([0.1, 0.2, 0.4, 0.8, 0.9])
ir = get_isotone_regression(X, y)
# Predictions should be monotonically increasing
test_x = np.linspace(0.5, 0.9, 10)
preds = ir.predict(test_x)
assert all(preds[i] <= preds[i+1] for i in range(len(preds)-1))
class TestHierarchicalLoss:
"""Tests for SCOPe hierarchical loss function."""
def test_exact_match(self):
"""Test exact match returns loss=0, exact=True."""
loss, exact = scope_hierarchical_loss('a.1.1.1', 'a.1.1.1')
assert loss == 0
assert exact is True
def test_family_mismatch(self):
"""Test family mismatch (last level) returns loss=1."""
loss, exact = scope_hierarchical_loss('a.1.1.1', 'a.1.1.2')
assert loss == 1
assert exact is False
def test_superfamily_mismatch(self):
"""Test superfamily mismatch returns loss=2."""
loss, exact = scope_hierarchical_loss('a.1.1.1', 'a.1.2.1')
assert loss == 2
assert exact is False
def test_fold_mismatch(self):
"""Test fold mismatch returns loss=3."""
loss, exact = scope_hierarchical_loss('a.1.1.1', 'a.2.1.1')
assert loss == 3
assert exact is False
def test_class_mismatch(self):
"""Test class mismatch returns loss=4."""
loss, exact = scope_hierarchical_loss('a.1.1.1', 'b.1.1.1')
assert loss == 4
assert exact is False
class TestValidation:
"""Tests for validation functions."""
def test_validate_lhat_new_returns_metrics(self, scope_like_data):
"""Test that validate_lhat_new returns expected metrics."""
sims, labels_exact = scope_like_data
labels_partial = labels_exact.copy() # Use same for simplicity
lhat = 0.9995 # Some threshold
error, frac_inexact, error_partial, frac_partial, fpr = validate_lhat_new(
sims, labels_partial, labels_exact, lhat
)
# All metrics should be in [0, 1]
assert 0 <= error <= 1
assert 0 <= frac_inexact <= 1
assert 0 <= error_partial <= 1
assert 0 <= frac_partial <= 1
assert 0 <= fpr <= 1
class TestIntegration:
"""Integration tests combining multiple components."""
def test_full_fdr_pipeline(self, calibration_test_split):
"""Test complete FDR control pipeline: calibrate → threshold → validate."""
data = calibration_test_split
alpha = 0.1
# Step 1: Get threshold from calibration data
lhat = get_thresh_new_FDR(
data['cal_sims'],
data['cal_labels'],
alpha
)
# Step 2: Calculate risk on test data
test_fdr = risk(data['test_sims'], data['test_labels'], lhat)
# FDR should be controlled (may be higher due to randomness in small samples)
# In practice with enough data, test_fdr should be <= alpha
assert test_fdr >= 0 # At minimum, should be valid
def test_full_fnr_pipeline(self, calibration_test_split):
"""Test complete FNR control pipeline."""
data = calibration_test_split
alpha = 0.1
# Get threshold for FNR control
lhat = get_thresh_new(
data['cal_sims'],
data['cal_labels'],
alpha
)
# Calculate FNR on test data
test_fnr = calculate_false_negatives(
data['test_sims'],
data['test_labels'],
lhat
)
# FNR should be controlled
assert test_fnr >= 0 # At minimum, should be valid
|