Data Science newsletter – November 4, 2016

GROUP CURATION: N/A

What do you get if you cross aspirin with ibuprofen? Harvard chemistry professor Alán Aspuru-Guzik isn’t sure, but he’s trained software that could give him an answer by suggesting a molecular structure that combines properties of both drugs.

After my short piece, “A hard road to reproducibility,” appeared in Science, I received several emails and Twitter mentions asking for more specific tips — both about tools and documents we use in the group to train the team about reproducibility.

The opacity of machine learning isn’t just an academic problem. More and more places use the technology for everything from image recognition to medical diagnoses. All that decision­making is, by definition, unknowable—and that makes ­people uneasy. My friend Zeynep Tufekci, a sociologist, warns about “Moore’s law plus inscrutability.” Microsoft CEO Satya Nadella says we need “algorithmic accountability.”

We usually think of mutations as errors in our genes that will make us sick. But not all errors are bad, and some can even cancel out or suppress the fallout of those mutations known to cause disease. While little has been known about this process — called genetic suppression — that will soon change as University of Toronto researchers uncover the general rules behind it.

Teams led by Professors Brenda Andrews, Charles Boone and Frederick Roth of the Donnelly Centre and the Department of Molecular Genetics, in collaboration with Professor Chad Myers of the University of Minnesota-Twin Cities, have compiled the first comprehensive set of suppressive mutations in a cell, as reported in the latest issue of Science. The four researchers are members of the Genetic Networks program of the Canadian Institute for Advanced Research. Their findings could help explain how suppressive mutations combine with disease-causing mutations to soften the blow or even prevent disease.

The South Big Data Hub’s Program to Empower Partnerships with Industry (PEPI) pairs early career faculty and researchers throughout the South with Industry Partners and support their travel to make collaboration possible. The program is co-sponsored by the National Science Foundation (NSF), the Computing Community Consortium (CCC), UnitedHealthCare Group and McKesson Corporation. Through PEPI, the South Hub provided funding to support data-intensive fellowships with industry for early career faculty, research scientists, and postdocs. Each award provided the recipient with up to $15,000 of travel and salary support to pay for their full-time effort for 2 – 5 weeks working at the company site.

Conventional wisdom holds that a scientist’s best work is usually published mid-career, in the sweet spot after they have learned the ropes, but before administrative duties or thoughts of retirement encroach upon research. So is an aging academic with an underwhelming research career a lost cause?

That was a motivating question behind a recent study by Kellogg’s Dashun Wang. “Sometimes when I give talks, I say this is ‘the hope project,’” says Wang, an associate professor of management and organizations. It is hopeful because Wang and colleagues find that a scientist’s most-cited paper is equally likely to pop up at any point in her career.

If this year’s landscape shows anything, it’s that the impact of machine intelligence is already here. Almost every industry is already being affected, from agriculture to transportation. Every employee can use machine intelligence to become more productive with tools that exist today. Companies have at their disposal, for the first time, the full set of building blocks to begin embedding machine intelligence in their businesses.

Smoking leaves an “archaeological record” of the hundreds of DNA mutations it causes, scientists have discovered.

Having sequenced thousands of tumour genomes, they found a 20-a-day smoker would rack up an average of 150 mutations in every lung cell each year.

At Columbia University’s History Lab, social scientists and data scientists have conducted many experiments to discover patterns and anomalies in official secrecy in large collections of declassified documents. We joined with collaborators at Fundação Getulio Vargas in Brazil, Renato Souza and Flavio Coelho, to see whether we could use data science methods to classify State Department communications.

We had two goals: First, find out whether, and to what extent, being classified as “secret” or “confidential” has historically been random or predictable. Second, learn what is normal and what might be considered negligent in how officials manage large numbers of potentially sensitive communications.

In this article, I propose making the process of commercializing preclinical research more reproducible and transparent by basing it on a blockchain. This effort will rely on the blockchain for communication to carry out peer reviews and publicly report the results. The program will be discussed thoroughly in a later section. Let us begin by reviewing three major initiatives currently in place to enhance reproducibility.

The difference between Facebook and Twitter reflects the shift from social media to messaging, which is just the latest in a series of shifts in Silicon Valley. Messaging services are now eclipsing social media companies in terms of user numbers. Asian messaging services like LINE and WeChat have already proven that they can monetize—something social media companies struggled with for a long time. (Arguably, Twitter is still, ten years into its existence, struggling to monetize itself). The move to messaging, a platform that can be monetized, and away from the more-nebulous category of social media is perhaps part of a movement away from quixotic ideas about growth, and a renewed focus on business fundamentals and a path to profitability.

Airbnb created an internal Knowledge Repo, combining git version control and Markdown templates for reporting results. Airbnb recently open-sourced its Knowledge Repository Beta, seeking contributors to help move the project forward.

Data scientists and organizations considering a cloud-based platform for their big data analytics needs can now evaluate Microsoft’s Data Science Virtual Machine for several hours without paying a dime.

“You can launch a VM instance with just a few clicks and explore it fully—no credit cards or Azure subscriptions needed,” wrote Microsoft staffers Paul Shealy, a senior software engineer, and Barnam Bora, a program manager, in a blog post. “A test drive lasts eight hours, enough time for you to try several sample solutions or analyze your own dataset.”

We present a set of computing tools and techniques that every researcher can and should adopt. These recommendations synthesize inspiration from our own work, from the experiences of the thousands of people who have taken part in Software Carpentry and Data Carpentry workshops over the past six years, and from a variety of other guides. Unlike some other guides, our recommendations are aimed specifically at people who are new to research computing.

Lots of stuff, hard to describe.