Ai Music Tutorial
Fito Coulter
Essential and state-of-the-art, The Computer Music Tutorial, second edition is a singular text that introduces computer and electronic music, explains its motivations, and puts topics into context. Curtis Roads's step-by-step presentation orients musicians, engineers, scientists, and anyone else new to computer and electronic music.
The new edition continues to be the definitive tutorial on all aspects of computer music, including digital audio, signal processing, musical input devices, performance software, editing systems, algorithmic composition, MIDI, and psychoacoustics, but the second edition also reflects the enormous growth of the field since the book's original publication in 1996. New chapters cover up-to-date topics like virtual analog, pulsar synthesis, concatenative synthesis, spectrum analysis by atomic decomposition, Open Sound Control, spectrum editors, and instrument and patch editors. Exhaustively referenced and cross-referenced, the second edition adds hundreds of new figures and references to the original charts, diagrams, screen images, and photographs in order to explain basic concepts and terms.
New chapters: virtual analog, pulsar synthesis, concatenative synthesis, spectrum analysis by atomic decomposition, Open Sound Control, spectrum editors, instrument and patch editors, and an appendix on machine learning
Curtis Roads is Professor of Media Arts and Technology, with an affiliate appointment in Music, at the University of California, Santa Barbara. His previous books include Microsound and Composing Electronic Music: A New Aesthetic.
This vignette provides a walk through tutorial on how to useMuSiC to estimate cell type proportions from bulksequencing data based on multi-subject single cell data by reproducingthe analysis in MuSiC paper, now is published on NatureCommunications.
Multi-subject single cell expression obtained from single-cellRNA sequencing (scRNA-seq). The cell types of scRNA-seq arepre-determined. These serve as reference for estimating cell typeproportions of bulk data.
MuSiC utilizes cell-type specific gene expression from single-cellRNA sequencing (RNA-seq) data to characterize cell type compositionsfrom bulk RNA-seq data in complex tissues. By appropriate weighting ofgenes showing cross-subject and cross-cell consistency, MuSiC enablesthe transfer of cell type-specific gene expression information from onedataset to another.
Solid tissues often contain closely related cell types which leads tocollinearity. To deal with collinearity, MuSiC employs a tree-guidedprocedure that recursively zooms in on closely related cell types.Briefly, we first group similar cell types into the same cluster andestimate cluster proportions, then recursively repeat this procedurewithin each cluster.
Bioconductor base package providesExpressionSet class, which is a convenient data structureto hold expression data along with sample/feature annotation. Here weuse two ExpressionSet objects to handle the bulk and singlecell data respectively. The details of constructingExpressionSet can be found on thispage. Please see the answer of this Issue for a simpleguidance.
Instead of selecting marker genes, MuSiC gives weights to each gene.The weighting scheme is based on cross-subject variation: up-weigh geneswith low variation and down-weigh genes with high variation. Here wedemonstrate step by step with the human pancreas datasets.
The dataset from Fadista et al. (2014)contains raw read counts data from bulk RNA-seq of human pancreaticislets to study glucose metabolism in healthy and hyper-hypoglycemicconditions. For the purpose of this vignette, the dataset ispre-processed and made available on the datadownload page. In addition to read counts, this dataset alsocontains HbA1c levels, BMI, gender and age information for eachsubject.
The single cell data are from Segerstolpe etal. (2016), which constrains read counts for 25453 genes across2209 cells. Here we only include the 1097 cells from 6 healthy subjects.The read counts are available on the datadownload page, in the form of anSingleCellExperiment.
The deconvolution of 89 subjects from Fadistaet al. (2014) are preformed with bulk dataGSE50244.bulk.eset and single cell referenceEMTAB.eset. We constrained our estimation on 6 major celltypes: alpha, beta, delta, gamma, acinar and ductal, which make up over90% of the whole islet.
The estimated proportions are normalized to sum to 1 across includedcell types. Here we use GSE50244.bulk.eset as thebulk.eset input and EMTAB.eset assc.eset input. The clusters is specified ascellType while samples issampleID. As stated before, we only included 6 major celltypes as select.ct.
Solid tissues often contain closely related cell types, andcorrelation of gene expression between these cell types leads tocollinearity, making it difficult to resolve their relative proportionsin bulk data. To deal with collinearity, MuSiC employs a tree-guidedprocedure that recursively zooms in on closely related cell types.Briefly, we first group similar cell types into the same cluster andestimate cluster proportions, then recursively repeat this procedurewithin each cluster. At each recursion stage, we only use genes thathave low within-cluster variance, a.k.a. the cross-cell consistentgenes. This is critical as the mean expression estimates of genes withhigh variance are affected by the pervasive bias in cell capture ofscRNA-seq experiments, and thus cannot serve as reliable reference.
The dataset GEOentry (GSE81492) (see Beckerman et al.2017) contains raw RNA-seq and sample annotation data. For thepurpose of this vignette, we will use the read counts dataMousebulkeset.rds from the datadownload page.
In previous MuSiCestimation procedure, the first step is to produce design matrix,cross-subject mean of relative abundance, cross-subject variance ofrelative abundance and average library size from single cell reference.These are taken care of by the function music_basis. Theessential inputs of music_basisare:
The immune cells are clustered together and the kidney specific cellsare clustered together. Notice that DCT and PT are within the samehigh-level grouping. The cut-off is user determined. Here we cut 13 celltypes into 4 groups:
We then select genes that are differentially expressed within clusterC3 (Epithelial cells) and C4 (Immune cells),available on data download page. Functionmusic_prop.clusteris used for estimation with pre-clustering of cell types. The essentialinputs are the same as music_prop except two unique inputs:groups and group.markers. groupspasses the column name of higher-cluster in phenoData. The intra-clusterdifferentially expressed genes are passed bygroup.marker.
Due to the limited space of Github, we can only demomusic_prop.cluster with a subset of mouse kidney singlecell dataset. Therefore, the results might be different from the onepresented in the paper due to incomplete reference single celldataset.
Benchmark dataset is constructed by summing up single cell data fromXinT2D.eset. The artificial bulk data is constructedthrough function bulk_construct.The inputs are single cell dataset, cluster name(clusters), sample name (samples) and selectedcell type (select.ct). bulk_constructreturns a ExpressionSet of artificial bulk datasetBulk.counts and a matrix of real cell type countsnum.real.
These are amazing! I had been asked to do the decorations for the annual Book and Thimble dinner and these are perfect! Quick and easy to make, thank goodness, as I will be making around 150 of them. Thank you for the wonderful tutorial. Great for those of us who REALLY need good pictures to follow a tutorial.
I work at a library and am in charge of disposing of severely damaged books. A co-worker named Rose is retiring and I was trying to think of a unique gift for her. Came across this great site and am making a boquet from discarded books. Rose has a little dog named Peanuts so I will put some peanuts in a clear vase and put the roses in it. Thank you so much for an easy and clear tutorial that even a craft-challenged person can follow through to a successful result!
What am I doing wrong? Every time I try to cut the petal design, I keep getting just a bunch of little bits of paper instead of the flower with the whole in the middle like you have. I need some more clear, strict directions!
I am using this tutorial with comic books for my superhero themed wedding on Oct 23. Thank you so much for making an easy to follow tutorial!! I will try to figure out how to post pics of the bouquets once they are done!
I made my wedding bouquet based on this template. If i could figure out how to attach a picture, I would. We added jewel centers to the roses (bought ofline). They were VERY fragile, but everyone said how gorgeous it was. We used pages from the first book my husband ever read to me, and made them together. It was really special. Thank you so much for the excellent guide.
The music program is a two-year program whose aim is to study music primarily as a liberal art. We seek to know the elements of music and in so doing, to know ourselves as musical beings and to know the musical world in which we live.
The Freshman Chorus prepares students for the work of the Sophomore Music Tutorial. The preparation is two-fold: students learn (if they do not already know) the basics necessary for reading music notation and they learn to sing great choral pieces that illustrate the musical elements studied in the Sophomore Music Tutorial. Freshman Chorus meets once a week and provides freshmen with a common musical experience, regardless of their formal education in music.
The Sophomore Music Tutorial seeks to develop an understanding of music through attentive listening, close study of musical theory, and the analysis of works of music by Bach, Beethoven, Mozart, Palestrina, Stravinsky, and Schoenberg, among others. Students undertake a thorough investigation of the diatonic system, a study of melody, counterpoint, and harmony, and an investigation of rhythm in words as well as in notes. A music tutorial has one tutor and 13 to 16 students and meets three times per week. Two shorter sessions are for discussion and one longer session brings students together to sing.
c80f0f1006