Follow along on #GI2013.

Starting today, you will notice that we have redesigned our website. This redesign is the proof of our commitment to continuously improve your reading experience. Through the use of bigger, clearer fonts, larger featured images, clear navigation menu, and a layout that makes it easier for you to scan through article titles, we have sought to improve every aspect of the site and the result is a better reading, discovery, and navigation experience. We have also added an “Editor’s Picks” section which will feature articles you will not want to miss out on.

We know that you love the freedom to access your favorite content from any of your devices, be it your desktop, laptop, tablet or smartphone. Our website automatically detects your screen size and adapts the site’s contents to your screen so that you have the best possible experience – load technophilicmag.com from your smartphone or tablet and see for yourself.

Thank you for reading Technophilic Magazine and don’t forget to send us your articles.

The Technophilic Team.

CAREER PATH

When you started your Ph.D., did you know you would go to industry upon graduating?

No. In fact, I did a Ph.D. because I wanted to figure out whether I wanted to pursue a career in academia or industry. By the time I started my research work in the mid 1970’s, silicon integrated circuits (ICs) really took off and I was immediately attracted by that direction. By the time I was writing my Ph.D. thesis, I knew that I would probably pursue a career in industry.

My thesis was a very theoretical analysis of semiconductor device behavior and I wanted to know how to do IC design. One of the job offers I received was to do IC design at Bell Northern Research, so I went back from California to Ottawa, where I designed one of the components of the first digital switch machine.

What did you do after Bell-Northern?

Intel started a research center in Portland, Oregon and they recruited a lot of the research scientists from Bell-Northern Research, so I decided to join Intel where I got to work on their commercial products. I was there for about 6 years.

But in 1983 – 1984, Taiwan started to become a hub for the semiconductor industry and they encouraged people with that background to go back to Taiwan and start semiconductor companies. So I had a decision to make: do I stay at Intel doing advanced product design, or do I take a big risk and join an early-stage startup and be unsure of what the future has in store for me a year down the road?

I went for the startup, which was called Quasel, as Chief Engineer. We wanted to build dynamic memory (DRAM), but I was very naïve at the time and didn’t realize that there are cycles to any industry: By the time we could produce our components, DRAM chips were selling cheaper than we could produce ours, so the company didn’t succeed. This was in 1986.

After that, I gathered a small team and started a small company that moved away from commodity components such as DRAM chips (because it’s a difficult market) and instead, focused on programmable chips.

How did you go from that to becoming an investor?

Almost a year into that startup, one of the investors of Quasel, Acer CEO Stan Shih, offered to merge my company with Acer, where I ran the Acer R&D Labs. That’s when I started my corporate executive career. About six years later, I came back to Silicon Valley as the President and CEO of Acer America, where I also got exposure to the field of investment because I felt that it was a rising trend.

In 1997, we setup Acer Technology Ventures (ATV), a venture fund that invested in very early stage startups. That’s how I evolved from a corporate executive to an investor. And I’ve been an investor ever since.

What do you think made you successful?

I was willing to take risks because I wasn’t afraid of failure. I told myself “If I fail, I will have learned something”.

STARTUPS & VCs

How does venture capital work from an investor’s point of view?

If I raise $100M from other investors, usually that money has to be invested in 5 to 7 years; you can’t sit on the money. So you have to use up the money during that period.How successful you are depends on how good is the return you get. And the successful VCs start to get more money so their size gets bigger and bigger. Some large Silicon Valley funds are now half a billion dollars; that’s a huge amount of money.

That’s a good thing, right?

It depends. Each VC has so-called venture partners and a large VC may have 10 or 20 partners that look for companies to invest. But if you have $500M overall to invest, each person needs to invest $25M in a short period. So what happens is that you tend to not invest in companies that require small amounts of money, which is bad news for early-stage companies.

Since the VCs now generally look for very large deals, that has changed the original nature of the VC industry. Today, those early-stage startups are now picked up by angel investors or seed stage/early-stage funds. Those early-stage companies will also typically get more one-on-one mentorship. But the way people get funding is constantly evolving. Outside of VCs, there’s also crowdfunding websites such as Kickstarter that are good places to start.

What makes you invest in a startup?

Three things: idea, market and team. Beyond the idea, I consider whether it covers a reasonably large market. The other aspect I consider is the team. Are they passionate about what they do? Experience is not as critical for young entrepreneurs but most importantly, do they have the ability to learn quickly and build a solid team around them?

Do VCs fund new ideas or better approaches to the same idea?

Some investors prefer investing in companies that do things better, faster or cheaper, while others prefer companies who do things that have never been done before. I prefer the latter.

Do you prefer when the founder stays in the company once investment is made?

It really depends. There are two kinds of founders: those who are very technical and only want to see their idea come to fruition, after which they want to return to do more fundamental research. In that case, we can find someone who’s capable of managing the company. Then there are founders who want to expand their knowledge on how to build up their business. When you start, ideally you’ll have a founder who has a technical background and who will eventually become a capable CEO. That’s the ideal case.

But often, a company will be started by 2-3 individuals and among them, maybe one will be more technical and the other will be more business oriented. That person usually ends up running the company.

LOOKING BACK…

Have you ever invested in a company and later found out you made a big mistake?

Oh yes! Typically, VCs invest in a number of portfolio companies. Our business model is that, out of 10 companies, there will be a spectrum of success: a couple will be very successful, 2 or 3 will be good and the rest will fail.

And even the very successful ones are very rarely overnight successes. Companies often find themselves redefining their business model along the way, and a good VC always tries to help companies figure out a way to go around those challenges

Some undergraduate students are nearing graduation and aren’t sure whether they should go on to grad school or industry. Do you have any advice for them?

The opportunities of today are much broader than 30-40 years ago and I don’t think there is one definite path. I think what students need to do is to find out where their passions lie.

For example, if you’re interested in doing research and experimentation, there are two paths. One is to continue on the academic path by going to graduate school, while the other is to conduct research in industrial labs. Today, the latter opportunity is less of an option; back in my day, there were many research institutions like Bell Labs, RCA Labs and Xerox PARC who were dedicated to tackling research questions. But today, this research is mostly attached to large corporations and has become more targeted to commercial applications, although a few of them still do advanced research like Intel and Google.

But once you get into that industry environment, I think it’s difficult to go back to academia. On the other hand, you can do a postdoctoral fellowship and maybe even become an assistant professor, and you probably still have the chance to go back to industry.

How do you remember your time at McGill?

I always felt like the time I spent at McGill was one of the best moments of my life. I really enjoyed the city, the school, the people I met and the professors I worked with.

Later on in your life, you will find that what matters is not only what you learn in class and in exams, but also the people you meet and the culture you’re exposed to, because each of you come from diverse background and have different ways of thinking

Explore the interactive map at cablemap.info.

Tell us about your background before college.

I have a very odd background. I grew up in a very poor rural area, where education wasn’t really something that was promoted and then went into the military. Through that, I got into college and when I started, it was a very intellectual sort of exercise, trying to discover how smart I was and how far I could push myself. So the combination of CS and pure math was a very natural place to go to push myself.

You started out by studying Math and CS. What got you into biology?

My undergraduate degree wasn’t so challenging so I decided to do my graduate studies in pure math, which I view as one of the most conceptually difficult areas of study. I was going through that but I always had an applied bent, and in pure math, it’s doing math for math’s sake: it’s not encouraged to figure out whether what you work on would satisfy another area of study. So out of curiosity, I wanted to figure out how everything we’re doing fits together, why we’re here and so on.

At UCLA, once I got the Ph.D. candidacy in pure math, I made the jump to a bio/math dual Ph.D. program that had the right level of rigor. I didn’t want to be a mathematical biologist of the type that were very good mathematicians but didn’t have a very deep understanding of the problems in biology and how to design your own experiments. I wanted to grasp that intuition; I wanted to think like a biologist.

Are you a biologist who does computation or a mathematician doing biology?

I view myself as a biologist who is heavily computational. Mathematicians that I’ve worked with in the past would not respect at all what I’m doing now as being real math. That always stings me a little bit, because biology has classically not been a very quantitative science so the kind of math that we do (e.g. Bayesian network reconstruction) is difficult but it’s not what a mathematician would view as the hardest thing!

Did you go to industry immediately after your Ph.D.?

Yes. What I saw while finishing up my Ph.D. was this revolution around technologies like the gene chips (used to recognize DNA from samples being tested) and microarrays (used to measure gene expression levels). Many companies said they’d start generating massive scales of data, house them in big databases and mine them, and that was unheard of in biology.

So I was very interested in those technologies and looked around for how I could get access to them. Roche Biosciences was among the first to sign all these big deals with the companies who were making the technology, and they appreciated the fact that you needed someone much more mathematical to look at the data. So I joined Roche. It was perfect timing.

What attracted you to Roche?

They had access to technologies that none of the universities had because of the outrageous costs at the time. What also drove me to Roche were the big resources and the excitement of having to carry out the right experiments to show proof-of-concept.

Because I was one of the first to apply statistical analysis for gene chips, I gained a certain degree of fame doing that, which caught the attention of the heads of Roche. But all of a sudden I was spending 50% of my time in meetings and fighting for why doing this is important instead of actually doing the science.

What did you do next?

I started talking to Rosetta, a startup that focused on building the technologies behind gene chips. After a year and a half at Roche, I went to Rosetta because they were more focused on the science. Also, since it was a startup, there was no bureaucracy or politics.

About a year and half later, Rosetta was bought by Merck. They loved what I was doing and they invested very heavily in that arm for 5-6 years. We did lots of good science and published a lot of papers. By the time I left Merck, we were responsible for about half of all the new drug discovery programs, so we were also delivering on the business side.

Why did you leave Merck?

Merck also got limiting because, as we were learning more about building these Bayesian networks, we wanted to go to the next level. We told them what we thought the next step should be but the price tag was about a billion dollars. That was too expensive for any one company to fund, so they were thinking more along the lines of turning this area into a pre-competitive space, where companies would be able to share all the data between each other.

After a lot of discussion, the co-Founder of Rosetta and I left Merck to found SAGE Bionetworks, a non-for-profit research center in Seattle. We focused on open-access biology: how to facilitate sharing of big data, how to build models and validate them, and enabling others to interact with those models.

What did you do once at SAGE?

Now that SAGE was set in motion, I joined Pacific Biosciences (PacBio). The idea was that I would setup a new institute in the Bay Area that would focus more on data generation and model building. PacBio knew that going in, and they liked the idea of me spending 75% of my time doing research outside the company because it would cost too much to have that big of a research effort going on internally. And for 25% of the time, I would be the Chief Scientific Officer at PacBio.

Although we got offers from UCSF and Stanford to set up the institute there, we needed ~$100M to really make a go at that project and we were having trouble finding enough money to make that project more than just my lab and myself. As I expanded the search for money, I locked onto Mount Sinai because we found donors inclined to give the $100M to do this effort.

What do you like best about Mount Sinai?

Compared to Stanford and UCSF, Mount Sinai had a reduced bureaucracy. Here, there’s a CEO who runs the hospital and the medical school. It’s a command-and-control architecture that I’m used to from the business side, where it’s easier to see things get done than one where every decision needs a committee. And it is smack down in the middle of a medical center, which will allow us to impact decision making directly in the clinic. That was very attractive. Moving to the East Coast is not something I thought I would ever do but all the pieces fell together!

Do you have advice about choosing between academia and industry?

I’ve always had a foot in academia and another in industry. Before joining Mount Sinai, however, the heavier foot was always in industry. This is the first time my heavy foot is in academia. I’ve seen both worlds for a long time and I think what academia offers is the ability to be your own CEO, grow out your own program and even though there are funding issues, you have much greater flexibility than in a company. You can make the kind of partnerships you need to leverage what is happening in industry and I view that as a more favorite path.

But I will say that the industry path offers, especially to young investigators, clarity of purpose and focus. Going to a startup is an experience like no other. Unlike academia, where you’re able to float and your timelines aren’t so critical, in a biotech startup, you’re living six months to six months. You have money and you see the cliff of when the money will end and if you don’t meet milestones, you’re going off that cliff and you’ll have to fire half the people in the company. That drives you to form bonds and work as a team to accomplish things far bigger than you could ever do. What you learn from that is invaluable.

The other advantage that I learned at Merck is: If what you’re working on is in the critical path of a company, the scale of resources you can get to carry out your vision is an order or two of magnitude greater than what you can ever get funded to do in academia, especially if it’s something new and risky.

What are the privacy issues with DNA sequencing becoming more popular?

We always want to protect data that can personally identify us but there’s another component of that which is the expectation of privacy. For example, your social security number can identify who you are and you should have an expectation of privacy around that. On the other hand, there are things like your face, which can be used to identify you but you have no reasonable expectation of privacy around your face.

DNA used to be in the camp of the social security number: of course you want to keep your DNA protected because it defines who you are. What’s changing now, however, is that the technology is becoming so amazing that, in 10 years, sequencing your genome will become as easy as taking a photograph? When that happens, there will still be the personal identifiable issue but the expectation of privacy will go away, because how can you have expectation of privacy around something that is as easy as taking a photograph. That is the transition we’re in. Of course you can take steps to protect the information but there are limits to what kind of privacy you can expect. Educating the population and legislators about that is critically important. The next step is to make laws that prevent discrimination based on that data.

What’s the biggest change you’d like to see in biology?

If biology wants to go to the next level in achieving an understanding of all the complex things we see, it needs to become much more quantitative and information-driven. Biology has to become a more physics-like discipline.

If biologists don’t do that, they’ll become irrelevant when Google, Amazon and other computer science powerhouses come in and do it before them. They won’t wait for the biologists to give them permission to analyze that data so if biologists aren’t there to work with them, they’ll be supplanted. I don’t think that’s extreme when you consider competitions where solving a biological problem gives you a $20,000 or $50,000 prize. If you look at who’s on the top of the leaderboard, none of them are biologists. Last time I checked, the top of the leaderboard was an accountant from Australia who knows nothing about biology.

Scientists often speak of “sequencing the human genome”. Is there really a single genome we can use as reference?

You used the key word ‘reference’: the Human Genome Project (HGP) was said to have sequenced the human genome. Really, the more accurate phrase that should’ve been used was: we created a reference sequence of the human genome. You shouldn’t think of the product of the HGP as the sequence of a human being, because that’s not actually true. In fact, what the HGP produced was a sequence of all human chromosomes—roughly 3 billion letters in total. But any given human being has 6 billion letters: 3 billion from mom, 3 billion from dad.

So we have to distinguish a reference sequence, which is the hypothetical representation of the sequence of each human chromosome, from a personal genome sequence, which is the full representation of the two copies of each of your chromosomes.

Is the reference genome very different from our genomes?

In some ways, it is. A reference sequence only differs from the genome you got from your parents by about 1 in a 1000 bases, which means that the reference sequence represents what you’ll find in the human species at about 99.9%. On the one hand, that’s incredibly similar. But on the other hand, the richness of what we want to learn is in that 0.1%; that’s what we’re most interested in if we think about health and disease. Those differences are called genetic variants and they can confer risk for disease or give protective characteristics.

So generating the human genome sequence is the HGP’s attempt at providing a framework—a reference or starting point—for being able to understand sequence differences and correlate those to health, disease, drug response and so forth.

So we can’t simply compare the reference to someone’s genome and look for differences?

That’s right. You don’t want to ask the question “does my genome differ from the reference?” That’s too simple of a question. The question you want to ask is “given those variants at this particular place in the genome, have they ever been seen before?” If so, how often have they been seen?

We now have databases that not only list all the variants that exist but also tell us the frequency with which we see them. So just because you differ from the reference sequence doesn’t mean anything.

Is all this research happening because it’s becoming cheaper to sequence?

A very important aspect is indeed the cost of sequencing that is dropping precipitously. The first human genome sequence cost us about 3 billion dollars—best $3 billion ever spent. Now the cost of sequencing your entire genome is on the order of $10,000, so we’ve gone from a billion to $10,000 in about 8 years. That’s pretty good. But we’re motivated to do this primarily because we know we have to: We can’t just have 1 human genome sequence, we need a whole lot more.

Can we go down to $100?

We proposed $1,000 in 2003 and we thought we were crazy. We would love it to be cheaper and cheaper but the truth of the matter is that we shouldn’t lose sight of where we are now. The $1,000 is very cheap compared to the cost of understanding what it actually means.

I don’t think that much about the cost of genome sequencing because I think we will eventually coast to the $1,000, or even $100, genome. That’s not where the burden is. Right now, the grand challenge is understanding that sequence: if I handed you your genome sequence—the perfect complete 6 billion letters—you would have to invest a lot of money to understand it. As of now, we don’t yet know how to interpret all that data, so the challenge now really lies not in data generation but data analysis.

How different are two genomes?

You and I roughly differ by 3 to 5 million single nucleotides, and the great majority of those are completely innocent—they have no phenotypic consequences. A small subset of those do, but we have very little knowledge about how to sift through them. We can make the list of variants but we’re not yet at the point where we can identify which ones we need to focus on and what their effect on human health is. We now need to take these catalogues of variants and start attributing biological and clinical relevance to them—that’s the next decade. Maybe people like you will help us figure it out.

Have we spotted some variants that are responsible for, say, cancer?

Sure, but we’re still at the very tip of the iceberg. Cancer is a great example because there’s a lot of action: It’s very clear that we need to sequence (and are now sequencing) cancer genomes and cataloging things that come in, and here’s why:

You can take 100 tumor samples and analyze them under the microscope: They’ll all look the same. But when you sequence their genomes, you might see that 50 of them tend to have a certain set of variants (i.e. a signature) and maybe the remaining 50 will have another signature.

We might even correlate a signature with groups of people that respond poorly to therapy. It would be great if we knew this upfront because it means we wouldn’t have to put poor reponders through chemotherapy. Instead, we’d look for more appropriate treatments

Try it out at my.pulseenergy.com/mcgill/dashboard

In your talk, you mentioned that optical communication is becoming more practical for shorter distances. What has permitted this reduction?

In the past, optical fibers were ultrafine and very delicate strands glass that required careful installation by highly trained personnel and tight mechanical tolerances. The cost of such installations could only be justified for transoceanic telecommunication or similar long-haul communication.

Advances in optics have recently allowed thicker and more bendable optical fibers to carry data at rates of 10+Gb/s. That’s fast enough to transmit an entire blue-ray disc in under 30 seconds. Using a thicker fiber relaxes tolerances everywhere in the system, making fiber optic installation easier, cheaper, and more robust.

Why is it difficult to do optical communication at very small distances?

The challenge is to make optical links economical and practical for use over very small distances. Currently, inexpensive optics are capable of communicating at data rates up to around 14 Gb/s, with research progressing towards commercial systems at 28 Gb/s. However, the optoelectronics at either end of the links (i.e. components responsible for converting the data to & from electrical signals) are very similar to those used 20 years ago when optical communication was reserved for long-haul links. The cost of these optoelectronic components are limiting the application of optical communication in areas that are cost-sensitive.

One area of research in your lab is CMOS photodetectors. Why is using CMOS an improvement?

CMOS is clearly the technology of our age. It has given us an ability to mass produce high-performance transistors at such low cost that it has transformed the world.CMOS has not only advanced computer chips. Digital image sensors were, for a long time, manufactured using CCD technology. But what really made image sensors ubiquitous was the discovery that by embedding a small circuit alongside each pixel of the sensor, CMOS image sensors can have a quality comparable to CCD sensors. Today, CMOS image sensors and digital cameras are everywhere, and creative uses for them continue to emerge.

Similarly, CMOS photodetectors with GHz bandwidth will enable a whole new set of applications for optical communication with far-reaching impact on our modern information age. Not only will they make optical communication less expensive, but more importantly they will enable optical links to be mass produced and integrated seamlessly into computing, memory, and wireless technologies using nanoscale CMOS manufacturing technologies.

In your talk, you mentioned that it is difficult to put photodetectors on CMOS. Why is that?

CMOS technology has been refined over decades to facilitate the fabrication of very high performance transistors. Unfortunately, the requirements of high performance transistors conflict with the requirements of high performance photodetectors: Tiny transistors require very thin interfaces between n-type and p-type silicon, whereas photodectors perform better when these interfaces, called depletion regions, are thick enough to absorb all photons incident on the detector.

When photodetectors are made using narrow depletion regions, many photons penetrate right through the depletion region resulting in a slow persistent current that can obscure the received data.

What has been done to circumvent these problems?

Several labs are trying to develop new manufacturing technologies that will permit the manufacture of both high performance transistors AND photodetectors. Unfortunately, those approaches imply increased cost. Our approach is improve the performance of CMOS photodetectors by making clever use of the high performance transistors already available in today’s CMOS at no additional cost. This is analogous to the advances that permitted CMOS technology to revolutionize image sensors.

What’s the difference with light travelling on a chip and when it does in optical fiber?

Light is significantly attenuated as it propagates through electronic chips, even after only a few millimeters. Light can travel along optical fiber for kilometers with little or no appreciable attenuation.

What other projects do you work on?

Another project my lab (the Integrated Systems Laboratory) is currently working on is to improve the energy-efficiency of distributed supercomputing environments by targeting the interconnections within them. The total energy per year consumed by compute servers is 220 TWh, roughly 10% of which is attributable to I/O.

Hence, even research that improves I/O energy efficiency by only 1% in these installations yields a savings equivalent to the average electricity consumption of 20,000 homes. Our research promises improvements far exceeding 1%