Bringing over tests/test_molevol.py (#40)

tests/test_molevol.py

@@ -0,0 +1,133 @@
+import torch
+import netam.molevol as molevol
+from netam import framework
+
+from netam.sequences import (
+    nt_idx_tensor_of_str,
+    translate_sequence,
+    AA_STR_SORTED,
+    CODONS,
+    NT_STR_SORTED,
+)
+
+# These happen to be the same as some examples in test_models.py but that's fine.
+# If it was important that they were shared, we should put them in a conftest.py.
+ex_mut_probs = torch.tensor([0.01, 0.02, 0.03])
+ex_sub_probs = torch.tensor(
+    [[0.0, 0.3, 0.5, 0.2], [0.4, 0.0, 0.1, 0.5], [0.2, 0.3, 0.0, 0.5]]
+)
+ex_parent_codon_idxs = nt_idx_tensor_of_str("ACG")
+parent_nt_seq = "CAGGTGCAGCTGGTGGAG"  # QVQLVE
+weights_path = "data/shmple_weights/my_shmoof"
+
+
+def test_build_mutation_matrix():
+    correct_tensor = torch.tensor(
+        [
+            # probability of mutation to each nucleotide (first entry in the first row
+            # is probability of no mutation)
+            [0.99, 0.003, 0.005, 0.002],
+            [0.008, 0.98, 0.002, 0.01],
+            [0.006, 0.009, 0.97, 0.015],
+        ]
+    )
+
+    computed_tensor = molevol.build_mutation_matrices(
+        ex_parent_codon_idxs.unsqueeze(0),
+        ex_mut_probs.unsqueeze(0),
+        ex_sub_probs.unsqueeze(0),
+    ).squeeze()
+
+    assert torch.allclose(correct_tensor, computed_tensor)
+
+
+def test_neutral_aa_mut_probs():
+    # This is the probability of a mutation to a codon that translates to the
+    # same. In this case, ACG is the codon, and it's fourfold degenerate. Thus
+    # we just multiply the probability of A and C staying the same from the
+    # correct_tensor just above.
+    correct_tensor = torch.tensor([1 - 0.99 * 0.98])
+
+    computed_tensor = molevol.neutral_aa_mut_probs(
+        ex_parent_codon_idxs.unsqueeze(0),
+        ex_mut_probs.unsqueeze(0),
+        ex_sub_probs.unsqueeze(0),
+    ).squeeze()
+
+    assert torch.allclose(correct_tensor, computed_tensor)
+
+
+def test_normalize_sub_probs():
+    parent_idxs = nt_idx_tensor_of_str("AC")
+    sub_probs = torch.tensor([[0.2, 0.3, 0.4, 0.1], [0.1, 0.2, 0.3, 0.4]])
+
+    expected_normalized = torch.tensor(
+        [[0.0, 0.375, 0.5, 0.125], [0.125, 0.0, 0.375, 0.5]]
+    )
+    normalized_sub_probs = molevol.normalize_sub_probs(parent_idxs, sub_probs)
+
+    assert normalized_sub_probs.shape == (2, 4), "Result has incorrect shape"
+    assert torch.allclose(
+        normalized_sub_probs, expected_normalized
+    ), "Unexpected normalized values"
+
+
+def iterative_aaprob_of_mut_and_sub(parent_codon, mut_probs, sub_probs):
+    """
+    Original version of codon_to_aa_probabilities, used for testing.
+    """
+    aa_probs = {}
+    for aa in AA_STR_SORTED:
+        aa_probs[aa] = 0.0
+
+    # iterate through all possible child codons
+    for child_codon in CODONS:
+        try:
+            aa = translate_sequence(child_codon)
+        except ValueError:  # check for STOP codon
+            continue
+
+        # iterate through codon sites and compute total probability of potential child codon
+        child_prob = 1.0
+        for isite in range(3):
+            if parent_codon[isite] == child_codon[isite]:
+                child_prob *= 1.0 - mut_probs[isite]
+            else:
+                child_prob *= mut_probs[isite]
+                child_prob *= sub_probs[isite][NT_STR_SORTED.index(child_codon[isite])]
+
+        aa_probs[aa] += child_prob
+
+    # need renormalization factor so that amino acid probabilities sum to 1,
+    # since probabilities to STOP codon are dropped
+    psum = sum(aa_probs.values())
+
+    return torch.tensor([aa_probs[aa] / psum for aa in AA_STR_SORTED])
+
+
+def test_aaprob_of_mut_and_sub():
+    crepe_path = "data/cnn_joi_sml-shmoof_small"
+    crepe = framework.load_crepe(crepe_path)
+    [rates], [subs] = crepe([parent_nt_seq])
+    mut_probs = 1.0 - torch.exp(-torch.tensor(rates.squeeze()))
+    parent_codon = parent_nt_seq[0:3]
+    parent_codon_idxs = nt_idx_tensor_of_str(parent_codon)
+    codon_mut_probs = mut_probs[0:3]
+    codon_subs = torch.tensor(subs[0:3])
+
+    iterative_result = iterative_aaprob_of_mut_and_sub(
+        parent_codon, codon_mut_probs, codon_subs
+    )
+
+    parent_codon_idxs = parent_codon_idxs.unsqueeze(0)
+    codon_mut_probs = codon_mut_probs.unsqueeze(0)
+    codon_subs = codon_subs.unsqueeze(0)
+
+    assert torch.allclose(
+        iterative_result,
+        molevol.aaprob_of_mut_and_sub(
+            parent_codon_idxs,
+            codon_mut_probs,
+            codon_subs,
+        ).squeeze(),
+    )