Many Paths Lead to Discovery: Analogical Retrieval of Cancer Therapies

Abstract

Abstract. This paper addresses the issue of analogical inference, and its potential role as the mediator of new therapeutic discoveries, by using disjunction operators based on quantum connectives to combine many potential reasoning pathways into a single search expression. In it, we extend our previous work in which we developed an approach to analogical retrieval using the Predication-based Semantic Indexing (PSI) model, which encodes both concepts and the relationships between them in high-dimensional vector space. As in our previous work, we leverage the ability of PSI to infer predicate pathways connecting two example concepts, in this case comprising of known therapeutic relationships. For example, given that drug x TREATS disease z, we might infer the predicate pathway drug x INTERACTS WITH gene y ASSOCIATED WITH disease z, and use this pathway to search for drugs related to another disease in similar ways. As biological systems tend to be characterized by networks of relationships, we evaluate the ability of quantum-inspired operators to mediate inference and retrieval across multiple relations, by testing the ability of different approaches to recover known therapeutic relationships. In addition, we introduce a novel complex vector based implementation of PSI, based on Plate's Circular Holographic Reduced Representations, which we utilize for all experiments in addition to the binary vector based approach we have applied in our previous research.

Embedding of semantic predications

“…On account of its use of a VSA, HolE shares ancestry with both PSI and ESP, and is perhaps most closely related to ESP on account of its use of gradient descent during training. Though we have used the Binary Spatter Code as the VSA for the current work, we have evaluated HRR-based implementations of PSI in previous work [49, 41, 60], and anticipate developing HRR-based implementations of ESP in the future. Like other models emerging from this community, HolE differs from both ESP and PSI as it learns parameters for predicate representations.…”

“…It has been found that the accuracy of such predictions can be improved by combining multiple reasoning pathways to increase the breadth of the search [49], and extending the length of the pathways to increase search depth [50]. In the former case, this is accomplished by using the span of vectors to model logical disjunction (OR), following the approach developed in [51].…”

Identifying plausible adverse drug reactions using knowledge extracted from the literature

“…We define the disjunction of these five query vectors as a query subspace derived from them using a binary vector approximation [93] of the Gram–Schmidt orthonormalization procedure [94]. The length of the projection of some other vector in this subspace provides an estimate of vector-subspace similarity.…”

Reasoning with vectors: A continuous model for fast robust inference

“…We have applied PSI to knowledge extracted by SemRep to infer therapeutic relationships between pharmaceutical agents and human diseases (Cohen et al, 2012a,b,c), using an approach we call discovery-by-analogy . The idea underlying this approach is to constrain the search for potential treatments to those that are connected to the disease in question along reasoning pathways suggesting therapeutic relationships.…”

“…These reasoning pathways were combined using the quantum disjunction operator to create a compound search expression which was used to retrieve treatments connected to other diseases across one, or several of these pathways (Cohen et al, 2012c,b). Further improvements in performance were obtained by extending the length of the predicate pathways concerned to include popular triple-predicate pathways also (Cohen et al, 2012a), allowing for the recovery of around ten percent more of the held-out set within the top one percent of predictions across all cancer types. This was accomplished by creating second-order semantic vectors for diseases, as the superposition of the semantic vectors of concepts that occurred in a predication of predicate type ASSOCIATED WITH with the disease in question, and using these as the starting point for inference instead of the disease in question.…”