scholarly journals EternaBrain: Automated RNA design through move sets and strategies from an Internet-scale RNA videogame

2018 ◽  
Author(s):  
Rohan V. Koodli ◽  
Benjamin Keep ◽  
Katherine R. Coppess ◽  
Fernando Portela ◽  
Rhiju Das ◽  
...  

ABSTRACTEmerging RNA-based approaches to disease detection and gene therapy require RNA sequences that fold into specific base-pairing patterns, but computational algorithms generally remain inadequate for these secondary structure design tasks. The Eterna project has crowdsourced RNA design to human video game players in the form of puzzles that reach extraordinary difficulty. Here, we demonstrate that Eterna participants’ moves and strategies can be leveraged to improve automated computational RNA design. We present an eternamoves-large repository consisting of 1.8 million of player moves on 12 of the most-played Eterna puzzles as well as an eternamoves-select repository of 30,477 moves from the top 72 players on a select set of more advanced puzzles. On eternamoves-select, we present a multilayer convolutional neural network (CNN) EternaBrain that achieves test accuracies of 51% and 34% in base prediction and location prediction, respectively, suggesting that top players’ moves are partially stereotyped. Pipelining this CNN’s move predictions with single-action-playout (SAP) of six strategies compiled by human players solves 61 out of 100 independent puzzles in the Eterna100 benchmark. EternaBrain-SAP outperforms previously published RNA design algorithms and achieves similar or better performance than a newer generation of deep learning methods, while being largely orthogonal to these other methods. Our study provides useful lessons for future efforts to achieve human-competitive performance with automated RNA design algorithms.

2021 ◽  
Author(s):  
Rohan V. Koodli ◽  
Boris Rudolfs ◽  
Hannah K. Wayment-Steele ◽  
Rhiju Das ◽  

AbstractThe rational design of RNA is becoming important for rapidly developing technologies in medicine and biochemistry. Recent work has led to the development of several RNA secondary structure design algorithms and corresponding benchmarks to evaluate their performance. However, the performance of these algorithms is linked to the nature of the underlying algorithms for predicting secondary structure from sequences. Here, we show that an online community of RNA design experts is capable of modifying an existing RNA secondary structure design benchmark (Eterna100) with minimal alterations to address changes in the folding engine used (Vienna 1.8 updated to Vienna 2.4). We tested this new Eterna100-V2 benchmark with five RNA design algorithms, and found that neural network-based methods exhibited reduced performance in the folding engine they were evaluated on in their respective papers. We investigated this discrepancy, and determined that structural features, previously classified as difficult, may be dependent on parameters inherent to the RNA energy function itself. These findings suggest that for optimal performance, future algorithms should focus on finding strategies capable of solving RNA secondary structure design benchmarks independently of the free energy benchmark used. Eterna100-V1 and Eterna100-V2 benchmarks and example solutions are freely available at https://github.com/eternagame/eterna100-benchmarking.


2020 ◽  
Author(s):  
Muhammad Nabeel Asim ◽  
Andreas Dengel ◽  
Sheraz Ahmed

ABSTRACTMicroRNAs are special RNA sequences containing 22 nucleotides and are capable of regulating almost 60% of highly complex mammalian transcriptome. Presently, there exists very limited approaches capable of visualizing miRNA locations inside cell to reveal the hidden pathways, and mechanisms behind miRNA functionality, transport, and biogenesis. State-of-the-art miRNA sub-cellular location prediction MIRLocatar approach makes use of sequence to sequence model along with pre-train k-mer embeddings. Existing pre-train k-mer embedding generation methodologies focus on the extraction of semantics of k-mers. In RNA sequences, rather than semantics, positional information of nucleotides is more important because distinct positions of four basic nucleotides actually define the functionality of RNA molecules. Considering the dynamicity and importance of nucleotides positions, instead of learning representation on the basis of k-mers semantics, we propose a novel kmerRP2vec feature representation approach that fuses positional information of k-mers to randomly initialized neural k-mer embeddings. Effectiveness of proposed feature representation approach is evaluated with two deep learning based convolutional neural network CNN and recurrent neural network RNN methodologies using 8 evaluation measures. Experimental results on a public benchmark miRNAsubloc dataset prove that proposed kmerRP2vec approach along with a simple CNN model outperforms state-of-the-art MirLocator approach with a significant margin of 18% and 19% in terms of precision and recall.


2014 ◽  
Vol 607 ◽  
pp. 118-123
Author(s):  
Lai Kuang Lin ◽  
Yi Min Xia ◽  
Fei He ◽  
Qing Song Mao ◽  
Kui Zhang

In view of complex and fuzziness of geological adaptive cutterhead selection for earth pressure balance (EPB) shield, a cutterhead selection method based on BP neural network is put forward. Considering the structure characteristics of EPB shield cutterhead, typical cutterhead types are classified and summarized based on cutterhead topology structure and number of spokes. After analyzing the determinants of cutterhead selection, one-to-many mapping relation between cutterhead type and geological parameters is put forward, and then core geologic parameters related to cutterhead selection are concluded. The feasibility of using neural network method to choose the cutterhead type is analyzed, and a BP neural network training model for cutterhead selection is set up and tested in testing sample data. The result shows that the selected cutterhead and the construction cutterhead are basically consistent. The feasibility of this method is proved and it can be theoretical basis for the cutterhead structure design which will improve scientific of cutterhead selection.


Sensors ◽  
2020 ◽  
Vol 20 (17) ◽  
pp. 4743
Author(s):  
Peisong He ◽  
Haoliang Li ◽  
Hongxia Wang ◽  
Ruimei Zhang

With the development of 3D rendering techniques, people can create photorealistic computer graphics (CG) easily with the advanced software, which is of great benefit to the video game and film industries. On the other hand, the abuse of CGs has threatened the integrity and authenticity of digital images. In the last decade, several detection methods of CGs have been proposed successfully. However, existing methods cannot provide reliable detection results for CGs with the small patch size and post-processing operations. To overcome the above-mentioned limitation, we proposed an attention-based dual-branch convolutional neural network (AD-CNN) to extract robust representations from fused color components. In pre-processing, raw RGB components and their blurred version with Gaussian low-pass filter are stacked together in channel-wise as the input for the AD-CNN, which aims to help the network learn more generalized patterns. The proposed AD-CNN starts with a dual-branch structure where two branches work in parallel and have the identical shallow CNN architecture, except that the first convolutional layer in each branch has various kernel sizes to exploit low-level forensics traces in multi-scale. The output features from each branch are jointly optimized by the attention-based fusion module which can assign the asymmetric weights to different branches automatically. Finally, the fused feature is fed into the following fully-connected layers to obtain final detection results. Comparative and self-analysis experiments have demonstrated the better detection capability and robustness of the proposed detection compared with other state-of-the-art methods under various experimental settings, especially for image patch with the small size and post-processing operations.


2011 ◽  
Vol 304 ◽  
pp. 18-23
Author(s):  
Chun Hua Hu

Resilient modulus of material is an important parameter for pavement structure design and analysis. However it is very tedious to get this parameter for hot mixture asphalt in laboratory. Moreover it takes long time to do experiments. In this paper, artificial neural network (ANN) is applied to predict to resilient modulus for hot mixture asphalt. A neural network model is constructed and trained plenty of times with selected test data until precision meets requirement. Then the model is used to predict resilient modulus for hot mix asphalt. Result of contrast prediction with test data shows that forecast precision is high. This provides a new method to predict resilient modulus for hot mixture asphalt.


Blood ◽  
2010 ◽  
Vol 116 (21) ◽  
pp. 3874-3874
Author(s):  
Joop Gaken ◽  
Azim Mohamedali ◽  
Jie Jiang ◽  
Farooq Malik ◽  
Austin G Kulasekararaj ◽  
...  

Abstract Abstract 3874 MicroRNAs are a class of small RNA molecules that regulate numerous critical cellular processes including proliferation, differentiation and apoptosis. Several microRNAs play important roles in normal haematopoiesis such as mir-181 which inhibits differentiation, mir-223 induces myeloid differentiation and mir-150 and mir-155 that are involved in T and B cell differentiation. Increased levels of mir-155 and mir-181 have been documented in diffuse large B-cell lymphomas and acute myeloid leukaemia respectively and mir-15a and mir-16 are frequently deleted in chronic lymphocytic leukaemia. More recently, down-regulation of mir-145 and mir-146a, located to the critical deleted region of 5q in MDS 5q- syndrome has been implicated in this disease. To obtain insight into the function of miRNAs, much effort has gone into different computational algorithms that identify miRNA targets. However, a major drawback of these prediction models is the substantial false positive rate and an inevitable bias due to reliance on the few known miRNA:target gene interactions. The lack of sensitivity and specificity of the developed computational algorithms is clearly shown by the fact that for the 940 human miRNAs identified (miRecords release 10 April 2010), only 152 miRNAs have experimentally validated targets. There is thus, a clear need to develop methodologies for the identification and validation of the functional targets of specific miRNAs. To enable the identification of biologically relevant microRNA targets we have developed a novel functional assay for the isolation of microRNA target sequences by selection, relying directly on downregulation, by a miRNA, of a selectable marker expressed in frame with a library of 3′ RNA sequences. The library was derived from human brain tissue cDNA, brain being the tissue expressing the largest number of individual genes (∼11,000). Cells with low or absent levels of the miRNA of interest are transfected with this 3′ UTR library inserted downstream of a TKzeo fusion gene in plasmid p3′TKzeo. Zeocin selection results in a population of cells that are expressing the TKzeo fusion protein and are resistant to zeocin and sensitive to Ganciclovir (GCV). The zeocin resistant cells are next transfected with the miRNA of interest cloned into the pbabepuro expression vector and selected in puromycin to isolate microRNA transduced cells. GCV treatment then selects for cells that have downregulated the TKzeo fusion protein expression either by inhibition of translation or mRNA cleavage. The 3′UTR sequences present downstream of the TKzeo fusion from GCV resistant cells are PCR amplified and sequenced. As proof of concept we identified targets for mir-130a which is involved in megakaryopoeisis and for which validated targets have been identified. This microRNA is not expressed in MCF7 cells, therefore the library was introduced into MCF7 cells and selected in 500μg/ml zeocin. Introduction of mir-130a, GCV selection, PCR amplification, cloning and sequencing of the introduced 3′ UTR sequences resulted in the identification of musculoaponeurotic fibrosarcoma oncogene homolog B (MAFB), a known validated target for mir-130a. In addition, we identified tumour protein translationally-controlled 1 (TPT1), proline rich 14 (PRR14), kinesin-associated protein 3 (KIFAP3), microtubule interacting and transport, domain containing 1 (MITD1) and cytochrome P450, family 27, subfamily A, polypeptide 1 (CYP27A1). All targets were validated by western blot analysis of cell extract from cells over-expressing mir-130a or cells treated with hairpin inhibitors directed against mir-130a. Deep sequencing of the total PCR product identified 107 putative targets for mir-130a which are being verified by western blot analysis of the genes for which antibodies are available. This strategy makes no assumptions based on previously identified sequences, relying directly on downregulation, by a miRNA, of a selectable marker expressed in frame with a library of 3′ RNA sequences. This strategy will lead to identification of functional targets for all the majority of microRNAs. For example the microRNAs present on the critical deleted chromosomal region of a number of haematological malignancies, including those on chromosomes 5 and 7 such as mir-143, mir-145, mir-146a, mir-378 etc. Disclosures: Gaken: Sigma: Patents & Royalties. Mohamedali:sigma: Patents & Royalties.


Nordlit ◽  
2019 ◽  
Author(s):  
Jaroslav Švelch

The article explores the manufacturing of monsters in video games, using the case of the influential 2007 first-person shooter BioShock, and ‘splicers’—its most numerous, zombie-like enemies. I combine two methodological perspectives on the ‘manufacturing’ of splicers by analyzing [a] the title’s developer commentary and other official paratexts to trace the design of splicers, and [b] the game’s embedded narrative to reconstruct the diegetic backstory of splicers. I argue that video game enemies, including splicers, are ‘computational others’, who may appear human on the level of representation, but whose behavior is machinic, and driven by computational algorithms. To justify the paradoxical relationship between their human-like representation and machinic behavior, BioShock includes an elaborate narrative that explains how the citizens of the underwater city of Rapture were dehumanized and transformed into hostile splicers. The narrative of dehumanization, explored following Haslam’s dehumanization theory (2006), includes [a] transforming splicers into atomized creatures by depriving them of political power and social bonds, [b] creating fungible and interchangeable enemies through splicers’ masks and bodily disintegration, [c] justifying splicers’ blindness to context and their simplistic behavior by portraying them as mentally unstable addicts. The article concludes that all video game enemies are inherently monstrous, and that critique of video game representation should focus on how games fail to make monsters human, rather than how games render humans monstrous or dehumanized.


Sign in / Sign up

Export Citation Format

Share Document