Merge branch 'main' into readme

Author: fkiraly

pyaptamer/datasets/tests/__init__.py

@@ -0,0 +1 @@
+"""Tests for datasets and their loaders."""

pyaptamer/experiments/__init__.py

@@ -0,0 +1,5 @@
+"""Base classes for experiments."""
+
+__all__ = ["Aptamer"]
+
+from pyaptamer.experiments._aptamer import Aptamer

pyaptamer/experiments/_aptamer.py

@@ -0,0 +1,113 @@
+__author__ = ["nennomp"]
+__all__ = ["Aptamer"]
+
+import torch
+from skbase.base import BaseObject
+
+from pyaptamer.utils import rna2vec
+
+
+class Aptamer(BaseObject):
+    """Candidate aptamer evaluation for a given target protein.
+
+    Parameters
+    ----------
+    target_encoded : torch.Tensor
+        Encoded target sequence tensor.
+    target : str, optional
+        Target sequence string.
+    model : torch.nn.Module
+        Model to use for assigning scores.
+    device : torch.device
+        Device to run the model on.
+
+    Examples
+    --------
+    from pyaptamer.experiment import Aptamer
+    >>> experiment = Aptamer(target_encoded, target, model, device)
+    >>> score = experiment.run(aptamer_candidate)
+    """
+
+    def __init__(
+        self,
+        target_encoded: torch.Tensor,
+        target: str,
+        model: torch.nn.Module,
+        device: torch.device,
+    ) -> None:
+        """
+        Parameters
+        ----------
+        target_encoded : torch.Tensor
+            Encoded target sequence tensor.
+        target : str, optional
+            Target sequence string.
+        model : torch.nn.Module
+            Model to use for assigning scores.
+        device : torch.device
+            Device to run the model on.
+        """
+        self.target_encoded = target_encoded.to(device)
+        self.target = target
+        self.model = model
+        self.device = device
+
+        super().__init__()
+
+    def _inputnames(self) -> list[str]:
+        """Return the inputs of the experiment."""
+        return ["aptamer_candidate"]
+
+    def _reconstruct(self, sequence: str = "") -> torch.Tensor:
+        """Reconstruct the actual aptamer sequence from the encoded representation.
+
+        The encoding uses pairs like 'A_' (add A to left) and '_A' (add A to right).
+        This method converts these pairs back to the actual sequence. Then, from its
+        RNA sequence representation it is converted to a vector.
+
+        Parameters
+        ----------
+        seq : str
+            Encoded sequence with direction markers (underscores).
+
+        Returns
+        -------
+        torch.Tensor
+            The reconstructed RNA sequence as a vector.
+        """
+        result = ""
+        for i in range(0, len(sequence), 2):
+            match sequence[i]:
+                case "_":
+                    # append the next values
+                    result = result + sequence[i + 1]
+                case _:
+                    # prepend the current value
+                    result = sequence[i] + result
+
+        return torch.tensor(rna2vec([result]))
+
+    @torch.no_grad()
+    def evaluate(self, aptamer_candidate: str) -> None:
+        """Evaluate the given aptamer candidate by assigning a score.
+
+        Parameters
+        ----------
+        aptamer_candidate : str
+            The aptamer candidate to evaluate. It should be a string consisting of
+            letters representing nucleotides: 'A_', '_A', 'C_', '_C', 'G_', '_G', 'U_',
+            '_U'. Underscores indicate whether the nucleotides are supposed to be (e.
+            g., 'A_') prepended or appended (e.g., '_A)'to the sequence.
+
+        Returns
+        -------
+        torch.Tensor
+            The score assigned to the aptamer candidate.
+        """
+        aptamer_candidate = self._reconstruct(aptamer_candidate)
+
+        self.model.eval()
+        return self.model(
+            aptamer_candidate.to(self.device),
+            self.target_encoded,
+        )

pyaptamer/mcts/__init__.py

@@ -0,0 +1,5 @@
+"""Monte Carlo Tree Search (MCTS) algorithm for string optimization."""
+
+__all__ = ["MCTS"]
+
+from pyaptamer.mcts._algorithm import MCTS