Feature attribution XAI algorithms enable their users to gain insight into the underlying patterns of large datasets through their feature importance calculation. Existing feature attribution algorithms treat all features in a dataset homogeneously, which may lead to misinterpretation of consequences of changing feature values. In this work, we consider partitioning features into controllable and uncontrollable parts and propose the Controllable fActor Feature Attribution (CAFA) approach to compute the relative importance of controllable features. We carried out experiments applying CAFA to two existing datasets and our own COVID-19 non-pharmaceutical control measures dataset. Experimental results show that with CAFA, we are able to exclude influences from uncontrollable features in our explanation while keeping the full dataset for prediction.

BACKGROUND: Spliced leader (SL) trans-splicing replaces the 5' end of pre-mRNAs with the spliced leader, an exon derived from a specialised non-coding RNA originating from elsewhere in the genome. This process is essential for resolving polycistronic pre-mRNAs produced by eukaryotic operons into monocistronic transcripts. SL trans-splicing and operons may have independently evolved multiple times throughout Eukarya, yet our understanding of these phenomena is limited to only a few well-characterised organisms, most notably C. elegans and trypanosomes. The primary barrier to systematic discovery and characterisation of SL trans-splicing and operons is the lack of computational tools for exploiting the surge of transcriptomic and genomic resources for a wide range of eukaryotes. RESULTS: Here we present two novel pipelines that automate the discovery of SLs and the prediction of operons in eukaryotic genomes from RNA-Seq data. SLIDR assembles putative SLs from 5' read tails present after read alignment to a reference genome or transcriptome, which are then verified by interrogating corresponding SL RNA genes for sequence motifs expected in bona fide SL RNA molecules. SLOPPR identifies RNA-Seq reads that contain a given 5' SL sequence, quantifies genome-wide SL trans-splicing events and predicts operons via distinct patterns of SL trans-splicing events across adjacent genes. We tested both pipelines with organisms known to carry out SL trans-splicing and organise their genes into operons, and demonstrate that (1) SLIDR correctly detects expected SLs and often discovers novel SL variants; (2) SLOPPR correctly identifies functionally specialised SLs, correctly predicts known operons and detects plausible novel operons. CONCLUSIONS: SLIDR and SLOPPR are flexible tools that will accelerate research into the evolutionary dynamics of SL trans-splicing and operons throughout Eukarya and improve gene discovery and annotation for a wide range of eukaryotic genomes. Both pipelines are implemented in Bash and R and are built upon readily available software commonly installed on most bioinformatics servers. Biological insight can be gleaned even from sparse, low-coverage datasets, implying that an untapped wealth of information can be retrieved from existing RNA-Seq datasets as well as from novel full-isoform sequencing protocols as they become more widely available.

The Society for the Study of Artificial Intelligence and Simulation of Behaviour (AISB) was founded in 1964 as the British Computer Society (BCS) Study Group on Artificial Intelligence and Simulati...

We introduce a layered approach to multi-agent programming and motivate this with a perspective to smart home environments. Apart from the core layer, layer components can be updated at runtime to reflect, e.g., attributes like credibility and the addition of proprietary functionality. The Layered Agent Framework (LAF) is defined by interfaces and organised into layers. This approach minimises system fragmentation while allowing developers to create and maintain meaningful and effective agents. A Petri net model is provided to visualise and execute prototypes of the agents. Although fully functional, the Petri nets will later be translated into dedicated programs with a smaller footprint and more efficient execution.

The aim of the LAM series of workshops is to bring together active researchers in the areas of logics and other formal frameworks for the modeling and analysis of dynamic or mobile systems.Many different approaches exist, some of which focus on resources, others focus on locations and locatedness, and yet others focus on communication.Coverage ranges from mathematical foundations to the practical analysis of protocols.The papers presented at the LAM workshops reflected this diversity.Consequently, the selected papers for this issue cover a wide range of topics: multi-modal logics, substructural semantics, π calculus and co-intuitionistic logics, proof theory, constraint automata, and object Petri nets.These topics are addressed in a variety of settings stretching from web programming to contextual information processing.

The verification and modelling of multi-agent systems is an important topic that has attracted much attention in recent years. Resources, however, have only recently been studied as simple extensions to well-known logics. Trying to find a set of useful features while retaining essential properties for practical use, we explore the question: Where are the limits of what can be verified about resource-bounded agents? We try to answer this question by considering several natural logic-based settings that may arise and prove that verification is usually undecidable apart from bounded or otherwise restrictive settings. Most interestingly, we identify various factors that influence the (un-)decidability and provide grounds for future research on more promising constraints leading to decidable fragments.

In this paper, we show that the flexible model-checking of multi-agent systems, implemented using agent-oriented programming languages, is viable thus paving the way for the construction of verifiably correct applications of autonomous agents and multi-agent systems. Model checking experiments were carried out on AJPF (agent JPF), our extension of Java PathFinder that incorporates the agent infrastructure layer, our unifying framework for agent programming languages. In our approach, properties are specified in a temporal language extended with (shallow) agent-related modalities. The framework then allows the verification of programs written in a variety of agent programming languages, thus removing the need for individual languages to implement their own verification framework. It even allows the verification of multi-agent systems comprised of agents developed in a variety of different (agent) programming languages. As an example, we also provide model checking results for the verification of a multi-agent system implementing a well-known task sharing protocol.

Abstract Object Petri nets (OPNs) provide a natural and modular method for the modelling of many real-world systems. We give a structure-preserving translation of OPNs to Prolog, avoiding the need for an unfolding to a flat Petri net. The translation provides support for reference and value semantics, and even allows different objects to be treated as copyable or non-copyable, respectively. The method is developed for OPNs with arbitrary nesting. We then apply logic programming tools to animate, compile and model check OPNs. In particular, we use the partial evaluation system logen to produce an OPN compiler, and we use the model checker xtl to verify CTL formulas. We also use logen to produce special purpose model checkers. We present two case studies, along with experimental results. A comparison to OPN translations to Maude specifications and model checking is given, showing that our approach is roughly twice as fast for larger systems. We also tackle infinite state model checking using the ecce system.

Object based Petri nets are becoming increasingly popular in many fields of computer science. The possibility to model real-world objects as separate Petri nets supports the need for modular design of complex systems. So far object net approaches have been based on the presumption that the object nets' structure remains unchanged in all processes. This paper sheds some light on possible extensions of high-level Petri nets to incorporate the dynamical evolution of Petri net structures. The exposition is based on the Linear Logic encoding of Petri nets ([1], [12]) and coloured Petri nets ([4]). It provides a basic semantics for modifying net structures which can be employed in a framework of nets within nets, i.e. situations where Petri nets (so-called token nets) themselves are used as tokens in an underlying environment net.

Linear Logic [4] has been shown to incorporate a fragment suitable for representing P/T-nets and giving a semantics to the computations of such nets (e.g. [1], [6], [2]). This result is generalised to coloured nets. Furthermore a new kind of high-lev