The application of natural language processing for the extraction of mechanistic information in toxicology.

To study the ways in which compounds can induce adverse effects, toxicologists have been constructing Adverse Outcome Pathways (AOPs). An AOP can be considered as a pragmatic tool to capture and visualize mechanisms underlying different types of toxicity inflicted by any kind of stressor, and describes the interactions between key entities that lead to the adverse outcome on multiple biological levels of organization. The construction or optimization of an AOP is a labor intensive process, which currently depends on the manual search, collection, reviewing and synthesis of available scientific literature. This process could however be largely facilitated using Natural Language Processing (NLP) to extract information contained in scientific literature in a systematic, objective, and rapid manner that would lead to greater accuracy and reproducibility. This would support researchers to invest their expertise in the substantive assessment of the AOPs by replacing the time spent on evidence gathering by a critical review of the data extracted by NLP. As case examples, we selected two frequent adversities observed in the liver: namely, cholestasis and steatosis denoting accumulation of bile and lipid, respectively. We used deep learning language models to recognize entities of interest in text and establish causal relationships between them. We demonstrate how an NLP pipeline combining Named Entity Recognition and a simple rules-based relationship extraction model helps screen compounds related to liver adversities in the literature, but also extract mechanistic information for how such adversities develop, from the molecular to the organismal level. Finally, we provide some perspectives opened by the recent progress in Large Language Models and how these could be used in the future. We propose this work brings two main contributions: 1) a proof-of-concept that NLP can support the extraction of information from text for modern toxicology and 2) a template open-source model for recognition of toxicological entities and extraction of their relationships. All resources are openly accessible via GitHub (https://github.com/ontox-project/en-tox).

Knowledge of collision modulates defensive multisensory responses to looming insects in arachnophobes.

We investigated the role of contextual knowledge in defensive responses to visual stimuli (spiders and butterflies) looming toward the hand. Human participants responded to tactile stimuli delivered to the same hand at 6 possible locations during an insect's approach. Tactile reaction times were faster when looming stimuli were closer to the hand, especially for spiders, and faster when insects loomed on a collision path than on a near-miss path. This latter finding suggests that human reactions to looming stimuli are not merely automatic reflexes but that contextual knowledge about the trajectory of looming objects is included in predicting their impact. (PsycInfo Database Record (c) 2022 APA, all rights reserved).

Natural language processing in toxicology: Delineating adverse outcome pathways and guiding the application of new approach methodologies.

Adverse Outcome Pathways (AOPs) are conceptual frameworks that tie an initial perturbation (molecular initiating event) to a phenotypic toxicological manifestation (adverse outcome), through a series of steps (key events). They provide therefore a standardized way to map and organize toxicological mechanistic information. As such, AOPs inform on key events underlying toxicity, thus supporting the development of New Approach Methodologies (NAMs), which aim to reduce the use of animal testing for toxicology purposes. However, the establishment of a novel AOP relies on the gathering of multiple streams of evidence and information, from available literature to knowledge databases. Often, this information is in the form of free text, also called unstructured text, which is not immediately digestible by a computer. This information is thus both tedious and increasingly time-consuming to process manually with the growing volume of data available. The advancement of machine learning provides alternative solutions to this challenge. To extract and organize information from relevant sources, it seems valuable to employ deep learning Natural Language Processing techniques. We review here some of the recent progress in the NLP field, and show how these techniques have already demonstrated value in the biomedical and toxicology areas. We also propose an approach to efficiently and reliably extract and combine relevant toxicological information from text. This data can be used to map underlying mechanisms that lead to toxicological effects and start building quantitative models, in particular AOPs, ultimately allowing animal-free human-based hazard and risk assessment.

When two worlds collide: the influence of an obstacle in peripersonal space on multisensory encoding.

Multisensory coding of the space surrounding our body, the peripersonal space, is crucial for motor control. Recently, it has been proposed that an important function of multisensory coding is that it allows anticipation of the tactile consequences of contact with a nearby object. Indeed, performing goal-directed actions (i.e. pointing and grasping) induces a continuous visuotactile remapping as a function of on-line sensorimotor requirements. Here, we investigated whether visuotactile remapping can be induced by obstacles, e.g. objects that are not the target of the grasping movement. In the current experiment, we used a cross-modal obstacle avoidance paradigm, in which participants reached past an obstacle to grasp a second object. Participants indicated the location of tactile targets delivered to the hand during the grasping movement, while a visual cue was sometimes presented simultaneously on the to-be-avoided object. The tactile and visual stimulation was triggered when the reaching hand passed a position that was drawn randomly from a continuous set of predetermined locations (between 0 and 200 mm depth at 5 mm intervals). We observed differences in visuotactile interaction during obstacle avoidance dependent on the location of the stimulation trigger: visual interference was enhanced for tactile stimulation that occurred when the hand was near the to-be-avoided object. We show that to-be-avoided obstacles, which are relevant for action but are not to-be-interacted with (as the terminus of an action), automatically evoke the tactile consequences of interaction. This shows that visuotactile remapping extends to obstacle avoidance and that this process is flexible.

Lateralized pointing does not cause a cognitive bias.

Lateralized pointing has been shown to cause not only a shift in visuo-motor midline, but also a shift in non-lateralized spatial attention. Non-lateralized cognitive consequences of lateralized pointing have been reported for local and global visuospatial processing. Here, we evaluate these findings and examine this effect for categorical and coordinate spatial relation processing, for which the attentional processes are thought to be highly similar to local and global visuospatial processing, respectively. Participants performed a commonly used working memory task to assess categorical and coordinate spatial relation processing. Lateralized pointing with either the left or the right hand, to either the left or the right side was introduced as a manipulation, as well as a new control condition without any pointing. Performance on the spatial relation task was measured before and after pointing. The results suggest that non-lateralized consequences of lateralized pointing cannot be generalized to other cognitive tasks relying on attentional processing. Further examination of lateralized pointing is recommended before drawing further conclusions concerning its impact on non-lateralized cognition.

No consistent cooling of the real hand in the rubber hand illusion.

In the rubber hand illusion (RHI), participants view a rubber hand that is stroked synchronously with their real, hidden hand. This procedure results in experiencing an increased sense of ownership over the rubber hand and demonstrates how multisensory information (vision, touch) can influence the sense of body ownership. However, it has also been suggested that a (lack of) sense of ownership over an own body part may in turn influence bodily processes. This suggestion has previously been supported by the observation that a decrease in skin temperature in the real hand correlated with ownership over the rubber hand. However, this finding has not been consistently replicated. Our lab has conducted several studies in which we recorded temperature of the hands during the RHI using various measures and in different circumstances, including continuous temperature measurements in a temperature-controlled room. An overall analysis of our results, covering five attempts to replicate the traditional RHI experiment and totalling 167 participants, does not show a reliable cooling of the real hand during the RHI. We discuss this failure to replicate and consider several possible explanations for inconsistencies between reports of hand temperature during the RHI.

Approaching threat modulates visuotactile interactions in peripersonal space.

The region surrounding our body (i.e. peripersonal space) is coded in a multimodal representation by fronto-parietal bimodal neurons integrating tactile stimuli on the body with nearby visual stimuli. This has often been suggested to serve a defensive purpose, which we propose could be mediated through visuotactile predictions. An approaching threat would then be of particular interest to peripersonal space processing. To investigate this, we asked participants to respond as fast as possible to a tactile stimulus on the hand, while looking at an animation of an approaching or receding spider or butterfly. Tactile stimulation was applied at one of 25 possible time points during the animation. Tactile reaction times were faster when an approaching stimulus was closer to the hand at the time of tactile presentation. Critically, this effect of distance on reaction times was larger when participants saw an approaching spider compared to an approaching butterfly, but only for participants who were afraid of spiders. This finding demonstrates that the perceived threat of an approaching stimulus modulates visuotactile interactions in peripersonal space and is consistent with the idea that visuotactile predictions are important for defensive purposes and maintaining bodily integrity.

Fingers crossed! An investigation of somatotopic representations using spatial directional judgements.

Processing of tactile stimuli requires both localising the stimuli on the body surface and combining this information with a representation of the current posture. When tactile stimuli are applied to crossed hands, the system first assumes a prototypical (e.g. uncrossed) positioning of the limbs. Remapping to include the crossed posture occurs within about 300 ms. Since fingers have been suggested to be represented in a mainly somatotopic reference frame we were interested in how the processing of tactile stimuli applied to the fingers would be affected by an unusual posture of the fingers. We asked participants to report the direction of movement of two tactile stimuli, applied successively to the crossed or uncrossed index and middle fingers of one hand at different inter-stimulus intervals (15 to 700 ms). Participants almost consistently reported perceiving the stimulus direction as opposite to what it was in the fingers crossed condition, even with SOAs of 700 ms, suggesting that on average they did not incorporate the unusual relative finger positions. Therefore our results are in agreement with the idea that, by default, the processing of tactile stimuli assumes a prototypical positioning of body parts. However, in contrast to what is generally found with tactile perception with crossed hands, performance did not improve with SOAs as long as 700 ms. This suggests that the localization of stimuli in a somatotopic reference and the integration of this representation with postural information are two separate processes that apply differently to the hands and fingers.

Thinking about touch facilitates tactile but not auditory processing.

Mental imagery is considered to be important for normal conscious experience. It is most frequently investigated in the visual, auditory and motor domain (imagination of movement), while the studies on tactile imagery (imagination of touch) are scarce. The current study investigated the effect of tactile and auditory imagery on the left/right discriminations of tactile and auditory stimuli. In line with our hypothesis, we observed that after tactile imagery, tactile stimuli were responded to faster as compared to auditory stimuli and vice versa. On average, tactile stimuli were responded to faster as compared to auditory stimuli, and stimuli in the imagery condition were on average responded to slower as compared to baseline performance (left/right discrimination without imagery assignment). The former is probably due to the spatial and somatotopic proximity of the fingers receiving the taps and the thumbs performing the response (button press), the latter to a dual task cost. Together, these results provide the first evidence of a behavioural effect of a tactile imagery assignment on the perception of real tactile stimuli.

An attentional approach to study mental representations of different parts of the hand.

The aim of this study is to investigate whether the fingers are represented separately from the palm. An exogenous spatial orientation paradigm was used where participants had to detect a tactile stimulus that could appear on the palm, the middle finger or the ring finger of the left hand. The tactile target was preceded by a non-predictive cue using different stimulus-onset asynchronies (SOA). We observed a Facilitation Effect in the palm and inhibition of return (IOR) for fingers using a short cue-target SOA, whereas the IOR was found in fingers and palm in long cue-target SOA. Also we observed a 'Cue above Target' effect (facilitation effect when the Cue had appeared distal to the target location in a vertical line) at the long SOA. Together, we suggest that the general pattern of results supports the proposed hypothesis about the different mental representation of fingers and palms, but with a considerable and hierarchical interrelation between them.