□ nick_lynch:
Good news @tonyhammond: New http://www.nature.com/ontologies RDF ontologies used by Macmillan science publishing #linkeddata
□ introspection:
Use and misuse of the gene ontology
annotations | @NatureRevGenet https://dpb.carnegiescience.edu/sites/dpb.carnegiescience.edu/files/rhee-etal-2008.pdf … #Bioinformatics via @R3RT0
More sophisticated approaches calculate the prob- ability of observing a particular enrichment value just by chance using a binomial model. The hypothesis-generating approach can also be valuable. For instance, an algorithm that predicts gene function on the basis of expression data should not include GO annotations based on microarray expression from those same experimental studies. Evidence codes and citations provided with each annotation can be used to filter annotations appropriately.
□ Seurat: Spatial reconstruction of single-cell gene expression data
>> http://www.nature.com/nbt/journal/vaop/ncurrent/full/nbt.3192.html#supplementary-information
a computational strategy to infer cellular localization by integrating single-cell RNA-seq data with in situ RNA patterns.
□ Complementary seminovaginal microbiome in couples
>> http://www.sciencedirect.com/science/article/pii/S0923250815000613 …
セックスメタゲノム、本当にあったのか…人間のパートナー同士の性交後の精液と膣内の比較
compare seminal and vaginal microbiomes in couples and to assess the influence of sexual intercourse on vaginal micro biome. Microbiomes of semen and vaginal fluid were profiled using Illumina HiSeq2000 sequencing of the V6 region of 16S rRNA gene.
□ Arvados Project Looks to New Models of Genomic Data Management:
>> http://www.bio-itworld.com/2015/4/14/arvados-project-looks-new-models-genomic-data-management.html
Different researchers put together their own pipelines of analysis tools to make sense of raw genomic data, and getting these pipelines to produce the same results in different compute environments is notoriously difficult
The vision that GA4GH have laid out, there will be a network of bioinformatics cores all over the world that will be storing genomic data, Arvados gets around this problem by letting users store stable copies of each pipeline they run, including an identical version of the dataset and every tool in the workflow.
“Our goal is to create infrastructure where you could take a million whole genomes and do machine learning at extremely high speeds, and you could do extremely complex genotyping queries in sub-seconds,”
□ Integrating alignment-based and alignment-free sequence similarity measures for biological sequence classification:
>> http://bioinformatics.oxfordjournals.org/content/early/2015/03/31/bioinformatics.btv006.full …
a new classification model that combines similarity scores obtained from alignment-free and alignment-based similarity measures with the aim to exploit the complementary nature of these measures to improve the classification accuracy. the CSSS method achieves a slightly better performance than the SW P value similarity/distance measure (the SVN or the 1-NN classifier) and performs much better than the combined LZW-BLAST similarity measure with the 1-NN classifier.
□ Wide-coverage relation extraction from MEDLINE using deep syntax:
>> http://www.biomedcentral.com/content/pdf/s12859-015-0538-8.pdf
a practical approach of creating PAS- based extraction patterns manually by observing actual linguistic expressions. no labeled corpora suitable for training a machine-learning based extraction model.
The linear-nonlinear-Poisson (LNP) encoding model formalizes the neural encoding process (cascade of three stages.)□ The Equivalence of Information-Theoretic and Likelihood-Based Methods for Neural Dimensionality Reduction:
>> http://journals.plos.org/ploscompbiol/article?id=10.1371/journal.pcbi.1004141
a common difficulty in the empirical estimation of information-theoretic quantities, and others working in more general machine-learning settings have suggested direct estimation of the ratio rather than its parts.
Let {rjt} denote spike counts during a “frozen noise” experiment, repeat index j Є {1,…,nrpt}, index t Є {1,…,nt} over time bins of width Δ.
□ District Data Labs - Modern Methods for Sentiment Analysis:
>> https://districtdatalabs.silvrback.com/modern-methods-for-sentiment-analysis
Architecture for Doc2Vec: DBOW predicts a random group of words in a paragraph given only its paragraph vector
from NNet import NeuralNet
nnet = NeuralNet(50, learn_rate=1e-2)
maxiter = 500
batch = 150
_ = nnet.fit(train_vecs, y_train, fine_tune=False, maxiter=maxiter, SGD=True, batch=batch, rho=0.9)
print 'Test Accuracy: %.2f'%nnet.score(test_vecs, y_test)
□ Towards a bioinformatics middle class - about dependencies & scripting vs. compiled languages
>> http://ivory.idyll.org/blog/2015-bioinformatics-middle-class.html
Fundamentally, moving from a lightweight Python layer on top of a heavier, optimized C++ library into a standalone binary
□ Nanopolish v0.2.0: HMM-based consensus caller for Oxford Nanopore data.
>> http://simpsonlab.github.io/2015/03/30/optimizing-hmm/ …
a quick hidden Markov model in Python to calculate the probability of observing a sequence of nanopore signals given an arbitrary sequence.
この人すごいな "I was not satisfied with the Python/C++ hybrid design. I admire Heng Li’s software where one usually just needs to run git clone.."
the calculation can be improved by using the transformation c=a+log(1+exp(b－a)) where a≥b. the forward algorithm of the HMM from 3,000μs per call to 278μs per call for 100 input events & a 100bp sequence, an improvment of over 10x.
# if ESL_LOG_SUM
return p7_FLogsum(a, b);
# else
if(a == -INFINITY && b == -INFINITY)
return -INFINITY;
if(a > b) {
double diff = b - a;
return a + log(1.0 + exp(diff));
} else {
double diff = a - b;
return b + log(1.0 + exp(diff));
}
□ Exploring Spark MLlib: Part 3 – Transformation and Model Creation:
>> https://phdata.io/exploring-spark-mllib-part-3-transformation-and-model-creation/ …
□ A simple data-adaptive probabilistic variant calling model:
>> http://www.ncbi.nlm.nih.gov/pmc/articles/PMC4363181/ …
a data adaptive model for variant calling based on easily accessible read characteristics, namely the log-likelihoods of nucleotide qualities relative read positions, alignment errors, multiple hits and the mismatch rate at a position to obtain a score. which are provided as input by the sequencing method, all log-likelihoods are sampled from the data itself.
□ Biological Dynamics Markup Language (BDML): an open format for representing quantitative biological dynamics data:
>> http://bioinformatics.oxfordjournals.org/content/31/7/1044.full
A limitation of the current version of BDML (0.2) is the lack of hierarchical representation of meta-information about genetic perturbations (e.g. mutants, gene editing and RNAi treatments) and chemical perturbations (e.g. drug treatments).
BioSignalML uses Resource Description Framework for encoding & storing of biomedical signals such as electrocardiograms & meta-information.
□ BioSignalML: An Abstract Model for Physiological Time-series Data:
>> https://researchspace.auckland.ac.nz/bitstream/handle/2292/22026/whole.pdf?sequence=2
BioSignalML annotation. Metadata mapping - Classes and properties from the BioSignalML Ontology & other vocabularies & ontologies are represented by class variables
□ StorageBIT: A Metadata-aware, Extensible, Semantic, and Hierarchical Database for Biosignals
>> http://www.it.pt/papconf_abs_p.asp?ID_PaperConference=12821&id=3
The Hierarchical Data Format is a self-describing data format designed to store and organize large amounts of numerical data.
a data model was defined and evaluated, database technolo- gies and file formats were presented, and various implementations were evaluated using a series of bench- marking tests for biosignal storage and retrieval, data insertion, update and querying. The results show that, first of all, MongoDB is, generally, faster than CouchDB as a DBMS.
’Biosignal NS’: {
’Duration ’
’Sample Rate ’
’Label ’: <Label [NS]> ’Transducer ’: <Transducer [NS]> ’Physical Dimension ’:
Mirroring of the EDF+ file structure onto the Data Model, using a JSON-style notation; text between ”
□ In between lines of code: On graph-based representations of a (set of) genomes:
>> https://flxlexblog.wordpress.com/2015/04/09/on-graph-based-representations-of-a-set-of-genomes/
a graph representing the genome of a species will grow (be updated) as more data on that species becomes available. Strategies to ‘lift over’ findings from earlier versions need to be developed
file formats to represent the graph are needed, but they would preferably be based on/compatible with existing community standards, or those standards should be easily derived from them
Two file formats to represent the graph been developed: fastg and GFA. Fastg has limited uptake, only two assembly programs (ALLPATHS_LG and SPAdes) will output in that format. GFA parsing is currently only experimentally in the ABYSS assembler, and also vg mentioned above is able to output it.
□ Two-group comparisons of zero-inflated intensity values: the choice of test statistic matters:
>> http://bioinformatics.oxfordjournals.org/content/early/2015/04/06/bioinformatics.btv154.short …
In the absence of distributional assumptions, the two-part Wilcoxon test or the empirical likelihood ratio test is recommended. For a vector of intensity values (including “zeros”) and a vector of group codes, R function calculates a likelihood ratio p-value and an estimate of the log fold change. direct estimates for the proportions of the two considered types of zero intensities (biological, technical) are provided.
□ Graphical algorithm for integration of genetic and biological data: proof of principle using psoriasis as a model:
>> http://bioinformatics.oxfordjournals.org/content/31/8/1243.short …
a novel approach, called Minimum distance-based Enrichment Analysis for Genetic Association (MEAGA) with the potential to address both of these important concerns. MEAGA computes a statistic summarizing the amount of overlapping genes and the overall shortest distance of the subgraph. and uses sampling strategy to approximate the null distribution of S and compute empirical and multiple testing-corrected p-values. MEAGA provides users copies of the compiled databases of shortest paths for interaction data obtained BioGrid, HPRD, and STRING.
../bin/MEAGA.py -s marker2gene.txt -g ../db/gene2fun.txt -i intmarkers -d ../db/splitFun_fungenesSP_BioGrid/ -o
□ GOBLET: The Global Organisation for Bioinformatics Learning, Education and Training:
>> http://journals.plos.org/ploscompbiol/article?id=10.1371/journal.pcbi.1004143 …
そういえば、こんな話あったなぁ。ファンディングモデルに興味あった。
the Outreach and PR (public relations) Committee is responsible for promoting GOBLET, maintaining its social networking interactions and galvanising communities to participate in its initiatives. GOBLET and ELIXIR communities to meet, to share training experiences and to discuss the scalable actions needed to implement a pan-European bioinformatics training strategy.
□ ELIXIREurope Innovation & SME Forum - data-driven growth in pharma biotech, Basel 9 June
>> http://goo.gl/QuUiaI @ISBSIB
□ "Don’t trust your data: reviewing Bioinformatics Data Skills"
>> http://www.molecularecologist.com/2015/04/dont-trust-your-data-reviewing-bioinformatics-data-skills/
learning bioinformatics by first, learning the general philosophy of working with data computationally. Instead of keeping the data in a set of ad hoc folders, plan your data management and store your work remotely often using version control.
bioinformaticsツールの再現性については、ドキュメント不足、他の実験では走らないなどの深刻なギャップが横たわったまま。最適化が乖離を生み出してるのか、相互のリテラシーのレベルの問題なのか
□ Simultaneous Discovery, Estimation and Prediction Analysis of Complex Traits Using a Bayesian Mixture Model
>> http://journals.plos.org/plosgenetics/article?id=10.1371/journal.pgen.1004969
a Bayesian mixture model and a priori assumed a mixture of four zero mean normal distributions of SNP effects (β),
estimate a hyper-parameter for the genetic variance from the data. compare BayesR with traditional single-SNP GWAS analyses, a linear mixed-effects modeling approach (LMM) and a Bayesian sparse linear mixed model (BSLM)
□ EBSeq-HMM: A Bayesian approach for identifying gene-expression changes in ordered RNA-seq experiments
>> http://bioinformatics.oxfordjournals.org/content/early/2015/04/03/bioinformatics.btv193.full.pdf
an empirical Bayes mixture modeling approach EBSeq-HMM has advantage in its ability to classify genes into particular expression paths. an auto-regressive hidden Markov model is implemented to accommodate dependence in gene expression across ordered conditions.
□ lncRScan-SVM: classify protein coding and long non-coding RNA (lncRNA) transcripts using support vector machine
>> http://lncrscansvm.sourceforge.net
□ sqawk: Sqawk is an Awk-like program that uses SQL and can combine data from multiple files. It is powered by SQLite.
>> https://github.com/dbohdan/sqawk
□ Notur: The Norwegian metacenter for computational science
>> https://www.notur.no/about
ノルウェーの計算科学の為の国家インフラ。UNINETT Sigmaが運営。これを利用したバイオインフォが面白いらしい
□ A hybrid model for a High-Performance Computing infrastructure for bioinformatics
>> https://flxlexblog.wordpress.com/2015/04/13/a-hybrid-model-for-a-high-performance-computing-infrastructure-for-bioinformatics/
"It is important to note that what we own is not located at our offices. Instead, these servers sit right next to the Abel servers, in the same rooms, sharing the same power and cooling setups, and even sharing the same disks. In other words, both our own servers and the Abel servers ‘see’ the same common disk areas!"
□ Deep Learning vs Probabilistic Graphical Models vs Logic:
>> http://quantombone.blogspot.jp/2015/04/deep-learning-vs-probabilistic.html
There is no reason why deep learning can't be combined with a GraphLab-style architecture, and some of the new exciting machine learning work in the next decade is likely to be a marriage of these two philosophies.
□ Semantic linking of complex properties, monitoring processes and facilities in representations of the environment
>> http://www.tandfonline.com/doi/abs/10.1080/17538947.2015.1033483#.VSSlJ0L2BE4
The presented ontology acknowledges that there are many complexities to the description of environmental properties which can be observed within the physical Earth system. The ontology is shown to be flexible and robust enough to describe concepts drawn from a range of Earth science disciplines, including ecology, geochemistry, hydrology and oceanography.
数年前から、水資源やグリッドに関するクラウドソーシングに参加したことはあるけど、この手の政治が関わる大規模問題に対し、一私企業にインセンティブを設ける方式では対応できないと痛感することが多かった。国際資金も含めオントロジーでOAにして、データを共有処理するのが第一歩というのが持論。
水資源に関するオントロジーデザインの一例。 ("A Semantic Portal for Next Generation Monitoring Systems")
In the area of ecological and environmental research, shallow integration approaches are taken to store and index metadata of data sources in a centralized database to aid search and discoverability.
□ Steven Weinbergの新著出てたのか。今度は歴史学批判で物議を醸してるようだけど、ワインバーグは昔からあんなスタンスよね。私も高校時代に"Dreams of a Final Theory"を読んで『哲学』の否定に影響を受けたからなぁ。純然たる理性と定量的手段を重んじる人
□ 因みに投資関連の私のバイブルは、Mauboussinの"Finding Financial Wisdom in Unconventional Places"だったりする。今でこそバズってる話だけど、投資科学と生物行動の知見は、こんなに自明ですよって何年も前に楔を打ち込んでる。
□ n0rr:
それがなんだかわからないものを扱っている我々にとっては、人間がシニフィエにシニフィアンを当てはめてやらなきゃいけない方法はよろしくないように思う。比較から得られた有意差それぞれに説明づけをしてくような絶望的作業になりそう。
□ 「生命の進化」が「◯◯する為に～」とか「◯◯という理由で～」という論説は20世紀にもう置いてきた方が良い害悪。進化という言葉自体に定向性を認めているから、生命事象に性格を付与するという根本的な過誤を犯している。
□ 「物事には全て理由がある」なんて言うけれど、これは実際には正反対の意味で、「結果には後付けで得られる説明しかない」ということの裏返しである。予測する手段を間違えないで運用することの方が重要だ。
□ GE and Veracyte are exploring the potential of combining their digital imaging and genomic technologies.
>> http://bit.ly/1NCUFtK
□ 仮説検定のN値とかp-hackingにしても、問題とされるのは作為性であり、有効範囲や規模、分野、時系列に依ってパラメータが動くから、成果物のデザイン自体を検証すべきという至極当然な帰結に至るしかない。そこだけ槍玉に挙げられてバズってる状況は結局の所、リテラシーの乖離が考えられる
□
