Chris Aldrich

Deep learning for computational biology | Molecular Systems Biology

Technological advances in genomics and imaging have led to an explosion of molecular and cellular profiling data from large numbers of samples. This rapid increase in biological data dimension and acquisition rate is challenging conventional analysis strategies. Modern machine learning methods, such as deep learning, promise to leverage very large data sets for finding hidden structure within them, and for making accurate predictions. In this review, we discuss applications of this new breed of analysis approaches in regulatory genomics and cellular imaging. We provide background of what deep learning is, and the settings in which it can be successfully applied to derive biological insights. In addition to presenting specific applications and providing tips for practical use, we also highlight possible pitfalls and limitations to guide computational biologists when and how to make the most use of this new technology.

Mol Syst Biol. (2016) 12: 878

Objectives

Over the past few years, an increasing number of large scale data sets have been made available in molecular biology. GTEx, for example, produced more than 18,000 RNA-Seq assays for multiple tissues in 900 individuals, Mindact generated gene expression data from about 7000 breast tumors in a single study, and 23andMe claims to have sequenced about 900,000 genomes. This growth in the available genomic data is expected to increase our capacity to identify cancer subtypes, regulatory genes, SNPs associated with phenotypes of interest, and biomarkers for many human traits. It also suggests exploring more complex feature representations when analyzing these datasets.

However, increasing the number of samples and features leads to a set of \textbf{interrelated statistical and computational problems}. Accordingly, the objectives of our workshop will be to:

Systematically identify the statistical and computational

problems arising during the analysis of large scale data in molecular biology;

Bring together experts in computational biology, molecular

biology, computer science, and statistics to propose innovative solutions to these problems, by leveraging recent advances in each of these fields.

Relevance, importance and timeliness

A number of studies generating high throughput molecular data for a large number of biological samples have been completed over the past five years. \textbf{Our workshop is important because the availability of these datasets holds great promises in terms of health improvement and understanding of molecular biology}. First of all, if exploited correctly, larger sample sizes should improve our ability to predict phenotypes of interest from molecular data. This entails very important applications such as improving the survival of cancer patients by better predicting which treatment they should receive, or decreasing bacterial resistances by predicting which antibiotic is efficient against a new strain. Correctly exploiting large scale datasets should also allow us to \textbf{better identify genetic and epigenetic determinants of these phenotypes, yielding a better understanding of human diseases and potentially guiding the development of new treatments and prevention policies}. In particular, more samples should allow the detection of less frequent variants in the human genome, or more complex features involving several modalities (copy number, expression, methylation, etc) associated with diseases. Finally, larger sample sizes should help with essential unsupervised tasks such as the \textbf{inference of regulation networks, or the identification of cancer subtypes}.

Our workshop is relevant because \textbf{all of these promises are conditioned on our solving of new statistical and computational challenges}. First (Challenge 1), we need to build new feature spaces and estimators whose complexity is adapted to these larger sample sizes, which involves designing novel, potentially more complex descriptors of the samples but still controlling the bias/variance trade-off. Second (Challenge 2), we need to build models which correctly integrate different modalities, such as copy number variation and gene expression. Third (Challenge 3), larger scale studies are more prone to unwanted variations, because they typically involve different labs and technical changes which can affect the measurements and become confounders in retrospective analyses. Similar or worse problems arise when trying to combine several existing datasets. We need methods which take this unwanted variation into account. Finally, (Challenge 4), we need new algorithms that make existing statistical tools scalable to the new sample sizes, and make estimation over the larger and more complex features of Challenge 1 tractable.

We also believe our workshop is very timely because \textbf{some of these statistical and computational challenges are starting to be addressed in other application fields} of statistics. It is crucial to recognize that the orders of magnitude are still very different in molecular biology and other data science application fields because of the cost and complexity of the data generation process: current large scale high throughput sequencing data sets typically contain a few thousand of samples but millions of features while computer vision, web, or astronomy datasets can involve billions or trillions of samples and relatively fewer features. A first consequence is that not all recent developments in machine learning are immediately transferable to computational biology. For example, so called deep learning methods have gained a lot of popularity and now represent the state of the art in computer vision but may not be the most appropriate tool for prediction of cancer outcome from molecular data. However, the fact that other fields already have much larger sample sizes also means that they had to develop efficient and scalable algorithms for basic tasks like feature selection, classification or clustering. \textbf{These recent developments are a great source of inspiration for computational biology, where large scale computation is still an emerging challenge}.

We believe \textbf{having a small scale workshop involving international experts in machine learning, statistics, computational biology and molecular biology is of utmost importance} for three main reasons. The first reason is that the technical advances we are referring to are very recent, often unknown to computational biologists and involve paradigms such as online optimization, accelerated gradient methods and network flow optimization, with which they are sometimes unfamiliar. The second reason is that it is not always obvious to non-statisticians which novel methods are appropriate given the current n/p regime. Conversely, the third reason is that statisticians do not know what the recent challenges are in molecular biology. Having them work on abstract versions of the problems is often not satisfactory as it is necessary to be aware of technical realities and of the underlying biology of the problem to come up with useful solutions.

My reply to ‘Novel, amazing, innovative’: positive words on the rise in science papers | Nature News & Comment
http://www.nature.com/news/novel-amazing-innovative-positive-words-on-the-rise-in-science-papers-1.19024

I can readily posit a potential and very strong motivation for the uptick in the appearance of these words, and particularly for the word "novel", which the paper and this article fail to see.

Beginning in the early 90's, and certainly starting before that time, there has been an intensified interest for both research institutions and researchers writing papers for them to monetize their research outside the halls of the academy. (Particularly with decreased funding for faculty and intensified competition for research dollars - many research faculty either self-fund their salaries or rely heavily on 3rd party income.) Toward this end, individual researchers began more aggressively pursuing patents for their research work and at the institutional level, colleges and universities began very actively pursuing technology transfer to the point of opening up full offices and hiring large numbers of staff to better leverage the return of creating patents and pursuing sales and licensing deals for the resulting research work. The US Patent Office, in making a determination of whether or not to grant patents, (and even moreso when the underlying documents are published scientific research articles) uses words like "novel" and "innovative" as an indicator of their worthiness. Without a patent, it's incredibly difficult to protect the intellectual property contained within a published and public work. Other words like "amazing" which are also cited in the paper are geared (in a business sales sense) more toward potential corporate financial investors who may consider purchasing or licensing the resulting work of a research paper. What corporate bean counter wants to invest in a dull-sounding research paper title which is quite likely one of the only parts of the document they're likely to comprehend? Yet throw in some "excitement words" and you may have a sale!

When one looks at employment contracts of the professoriate (post-docs, graduate students, and the like), which presumably comprises the majority of those publishing papers with these words, one will see that professors prior to the 90's have contracts that don't mention intellectual property rights resulting from their efforts, or which grant them the lion's share. Current contracts after that period will almost necessarily keep either all or most of these types of revenue streams in the institution's pockets rather than the researchers themselves. I'd be willing to bet that this tide began turning in the mid-1970s and has certainly been overwhelming since the mid-90s.

It is these severe economic taskmasters which are almost assuredly the cause of the rise of these buzzwords in research papers, and most particularly in their titles. One need only ask themselves, which university wants to be the next proverbial "Stanford" to not have a piece of revenue from the intellectual property of a proverbial "Page rank" algorithm? Or in biomedical research, to not take a hefty cut out of a drug which could potentially cure cancer or HIV/AIDS? Take a gander at what something as simple as a HELA cell, which was taken and cultured without any notice to Henrietta Lacks, has become from an industrial perspective. If they can avoid it, academia certainly won't let multi-million (or billion) dollar enterprises spring up (again) without them receiving a piece of the pie.

Naturally, there are other factors compounding the issues at hand, particularly when one looks at negative research findings, which are sadly rarely ever published, particularly in an era of big data when what we know isn't true will surely help assist in defining the boundaries of what we do know and what could potentially be determined at that boundary. (Much the way that white space in photographs or in printed pages helps to define an image as much if not more so than the darker portions of the image.) Negative words are likely more difficult to get published as it may reflect poorly on researchers who are reliant on continued future funding and may see their sources disappear when they're not producing groundbreaking work. My hypothesis, however, is that it's more likely a direct economic impact based on technology transfer than it is the increased competitive landscape in academic research.

<i>Manny wrote: "I see Jacobson was an expert on Lie algebras... yes, that might certainly be useful! Thanks..."</i>

Many/most graduate math texts on Lie Groups/Algebras utilize manifold theory as their basis, which can make the field more daunting for physicists. As a result, I recommend looking at books that take a more linear algebraic bent to the subject, which can make some of the quantum mechanics related areas more transparent. I've used [book:Lie Groups, Lie Algebras, and Representations: An Elementary Introduction|1051566], [book:Matrix Groups for Undergraduates.|1162374], and [book:Lie Groups, Lie Algebras, and Some of Their Applications|424103] for these types of applications and viewpoints.

About a year ago, I took a two quarter sequence at UCLA on Lie Groups from an ("easier") matrix group perspective which I imagine you may find somewhat useful in your reading on quantum mechanics. Several engineers, programmers, amateur mathematicians, physicists, and quantum mechanics enthusiasts had spent several years coaxing the professor into teaching it from this perspective. It was geared toward the advanced undergraduate level, and based on your comments here and your reading of Shankar, should be relatively easily followable.

We loosely followed Hall's textbook and portions of [book:Matrix Analysis|647523], which will give you some of the advanced linear algebra you could possibly be missing depending on your background. If it helps, here's a link to a downloadable pdf copy of the notes for the first class with the audio of the lecture embedded (using Livescribe.com digital pen technology which should let you click on the notes and jump to the audio portion related to where you've clicked): http://bit.ly/1KMxj0R. If it's useful, let me know and I can give you links for others. You may need to open it up in a more recent version of Acrobat Reader to be able to access the audio portion of the lecture, which will go a long way to assisting the clarity of the notes.)

Having this background may make Weyl and Woit's developing textbook more easily manageable. My guess is that Woit is doing a more thorough job of developing the math than typical physics-oriented texts like [book:Geometry, Topology and Physics|439357] which do a lot of hand-waving at the math in an effort to get to the physics more quickly, but at the detriment of understanding what is happening mathematically.

I've dipped into some of Weyl's work in the past, but also keep in mind that some of his notation and definitions can be dated in relation to more modern presentations.

Deep learning for computational biology | Molecular Systems Biology

17w5131: Statistical & Computational Challenges in Large Scale Molecular Biology Workshop @BIRS_Math 3/2017 #ITBio

Organizers

Objectives

Relevance, importance and timeliness