Background: Complex networks are studied across many fields of science and are particularly important to understand biological processes. Motifs in networks are small connected sub-graphs that occur significantly in higher frequencies than in random networks. They have recently gathered much attention as a useful concept to uncover structural design principles of complex networks. Existing algorithms for finding network motifs are extremely costly in CPU time and memory consumption and have practically restrictions on the size of motifs.

Reinforcement learning (RL) has recently regained popularity with major achievements such as beating the European game of Go champion. Here, for the first time, we show that RL can be used efficiently to train a spiking neural network (SNN) to perform object recognition in natural images without using an external classifier. We used a feedforward convolutional SNN and a temporal coding scheme where the most strongly activated neurons fire first, while less activated ones fire later, or not at all. In the highest layers, each neuron was assigned to an object category, and it was assumed that the stimulus category was the category of the first neuron to fire. If this assumption was correct, the neuron was rewarded, i.e., spike-timing-dependent plasticity (STDP) was applied, which reinforced the neuron's selectivity. Otherwise, anti-STDP was applied, which encouraged the neuron to learn something else. As demonstrated on various image data sets (Caltech, ETH-80, and NORB), this reward-modulated STDP (R-STDP) approach has extracted particularly discriminative visual features, whereas classic unsupervised STDP extracts any feature that consistently repeats. As a result, R-STDP has outperformed STDP on these data sets. Furthermore, R-STDP is suitable for online learning and can adapt to drastic changes such as label permutations. Finally, it is worth mentioning that both feature extraction and classification were done with spikes, using at most one spike per neuron. Thus, the network is hardware friendly and energy efficient.

The primate visual system has inspired the development of deep artificial neural networks, which have revolutionized the computer vision domain. Yet these networks are much less energy-efficient than their biological counterparts, and they are typically trained with backpropagation, which is extremely data-hungry. To address these limitations, we used a deep convolutional spiking neural network (DCSNN) and a latency-coding scheme. We trained it using a combination of spike-timingdependent plasticity (STDP) for the lower layers and reward-modulated STDP (R-STDP) for the higher ones. In short, with R-STDP a correct (resp. incorrect) decision leads to STDP (resp. anti-STDP). This approach led to an accuracy of 97.2% on MNIST, without requiring an external classifier. In addition, we demonstrated that R-STDP extracts features that are diagnostic for the task at hand, and discards the other ones, whereas STDP extracts any feature that repeats. Finally, our ap- * Corresponding author.Email addresses: milad.mozafari@ut.ac.ir (MM), mgtabesh@ut.ac.ir (MG) nowzari@ut.ac.ir (AND) simon.thorpe@cnrs.fr (SJT) timothee.masquelier@cnrs.fr (TM).proach is biologically plausible, hardware friendly, and energy-efficient.

Our goal of this study was to reconstruct a “genome-scale co-expression network” and find important modules in lung adenocarcinoma so that we could identify the genes involved in lung adenocarcinoma. We integrated gene mutation, GWAS, CGH, array-CGH and SNP array data in order to identify important genes and loci in genome-scale. Afterwards, on the basis of the identified genes a co-expression network was reconstructed from the co-expression data. The reconstructed network was named “genome-scale co-expression network”. As the next step, 23 key modules were disclosed through clustering. In this study a number of genes have been identified for the first time to be implicated in lung adenocarcinoma by analyzing the modules. The genes EGFR, PIK3CA, TAF15, XIAP, VAPB, Appl1, Rab5a, ARF4, CLPTM1L, SP4, ZNF124, LPP, FOXP1, SOX18, MSX2, NFE2L2, SMARCC1, TRA2B, CBX3, PRPF6, ATP6V1C1, MYBBP1A, MACF1, GRM2, TBXA2R, PRKAR2A, PTK2, PGF and MYO10 are among the genes that belong to modules 1 and 22. All these genes, being implicated in at least one of the phenomena, namely cell survival, proliferation and metastasis, have an over-expression pattern similar to that of EGFR. In few modules, the genes such as CCNA2 (Cyclin A2), CCNB2 (Cyclin B2), CDK1, CDK5, CDC27, CDCA5, CDCA8, ASPM, BUB1, KIF15, KIF2C, NEK2, NUSAP1, PRC1, SMC4, SYCE2, TFDP1, CDC42 and ARHGEF9 are present that play a crucial role in cell cycle progression. In addition to the mentioned genes, there are some other genes (i.e. DLGAP5, BIRC5, PSMD2, Src, TTK, SENP2, PSMD2, DOK2, FUS and etc.) in the modules.

Application of deep convolutional spiking neural networks (SNNs) to artificial intelligence (AI) tasks has recently gained a lot of interest since SNNs are hardware-friendly and energy-efficient. Unlike the non-spiking counterparts, most of the existing SNN simulation frameworks are not practically efficient enough for large-scale AI tasks. In this paper, we introduce SpykeTorch, an open-source high-speed simulation framework based on PyTorch. This framework simulates convolutional SNNs with at most one spike per neuron and the rank-order encoding scheme. In terms of learning rules, both spike-timing-dependent plasticity (STDP) and reward-modulated STDP (R-STDP) are implemented, but other rules could be implemented easily. Apart from the aforementioned properties, SpykeTorch is highly generic and capable of reproducing the results of various studies. Computations in the proposed framework are tensor-based and totally done by PyTorch functions, which in turn brings the ability of just-in-time optimization for running on CPUs, GPUs, or Multi-GPU platforms.

The strongly NP-Hard Double Digest Problem, for reconstructing the physical map of DNA sequence, in now using for efficient genotyping. Most of the existing methods are inefficient in tackling large instances due to the large search space for the problem which grows as a factorial function (a!)(b!) of the numbers a and b of the DNA fragments generated by the two restriction enzymes. Also, none of the existing methods are able to handle the erroneous data. In this paper, we develop a novel method based on genetic algorithm for solving this problem and it is adapted to handle the erroneous data. Our genetic algorithm is implemented and compared with the other well-known existing algorithms. The obtained results show the efficiency (speedup) of our algorithm with respect to the other methods, specially for erroneous data.

Stem cells, with their capacity to self-renew and to differentiate to more specialized cell types, play a key role to maintain homeostasis in adult tissues. To investigate how, in the dynamic stochastic environment of a tissue, non-genetic diversity and the precise balance between proliferation and differentiation are achieved, it is necessary to understand the molecular mechanisms of the stem cells in decision making process. By focusing on the impact of stochasticity, we proposed a computational model describing the regulatory circuitry as a tri-stable dynamical system to reveal the mechanism which orchestrate this balance. Our model explains how the distribution of noise in genes, linked to the cell regulatory networks, affects cell decision-making to maintain homeostatic state. The noise effect on tissue homeostasis is achieved by regulating the probability of differentiation and self-renewal through symmetric and/or asymmetric cell divisions. Our model reveals, when mutations due to the replication of DNA in stem cell division, are inevitable, how mutations contribute to either aging gradually or the development of cancer in a short period of time. Furthermore, our model sheds some light on the impact of more complex regulatory networks on the system robustness against perturbations.

No abstract