R and its rising role

Interview with Coline Zeballos

How does information technology help reduce the statistical margin of mistakes?
How does leadership come into play when creating programs designed to effectively manage statistical data?
How does open-source culture in informatics development in statistics help in creating innovative technology to increase research validity?
Does open-source development change the way statistical validation is treated, especially in developing new forms of health treatment?

My interview with Coline Zeballos of Roche informatics will give you an idea on how the development of Open-Source culture in developing applications that will help validate data collected for statistical research is actually changing the way medical studies are being completed and applied by pharma companies today.

Coline also shared her thoughts about internal leadership in this podcast, which will be useful not only for statisticians but for professionals in other industries as well.

Some lessons you can get from this podcast include: 

  • Using the R Strategy to improve the way teams work towards a singular focused goal.
  • Improving how teams are directed and delegated in tasks where they can perform best.
  • Using internal leadership alignment to help key persons become more effective in their assigned posts.

Head on to the podcast, listen to our conversation, learn more about effective developments in informatics and how they change the culture of statistics.

Share this with your peers and we hope to give you inspiration as we gear ourselves towards more efficient modern statistical developments now and in the future for better data utilization.

Coline Zeballos

R Strategy Lead at RocheProfile photo of Coline Zeballos

She has worked for GlaxoSmithKline Consumer Healthcare and Zurich Insurance Group, learned Data Science with Python for businesses, led the business development in Switzerland of BAMBOO Tec, a data analytics company based in Switzerland and Bolivia.

Today, she continues her professional journey supporting Roche in its major shift towards a more automated, efficient way of getting drugs on the market thanks to a language-agnostic mindset and set of tools for data scientists and biostatisticians.

What she enjoys the most is working with people from different backgrounds and bringing such interdisciplinary teams to their best performance. She especially appreciates making the link between data science and business needs. Her work is based on a strong collaboration between functions that she drives with subtlety, as well as a great ability to simplify complex environments.

Throughout the years, she has acquired skills such as agile business development, project management, team management, leadership, data science, interactive data visualization, accelerating decision-making with data analytics tools, and data storage.


[00:00:00] Alexander: You’re listening to The Effective Statistician Podcast, the weekly podcast with Alexander Schacht and Benjamin Piske. Designed to help you reach your potential lead great science and serve patients without becoming overwhelmed by work. Today I’m talking with Coline Zeballos who is a strategic person. Working on our at rush and success will be pretty cool because you’ll learn a lot about what is currently ongoing in that space and how it will impact our industry.

And also you’ll learn a little bit about strategy. By his way. So stay tuned for this really nice interview with Colleen.

I’m producing this podcast in association with psi, the community dedicated to leading, and promoting the use of statistics since the healthcare industry for the benefit of patients. Join PSI today to further develop your statistical capabilities with access to the video on demand content library free registration to all PSI have announced and much, much more.

Head over to PSI web dot log to learn more about PSI activities, DETERMINE PSI today.

Welcome to episode and today I’m really happy to have Colleen from Rush with us and we’ll speak about lots of stuff about art, which is, Pretty big change in the organization. And with that, there’s a lot of other associated changes. Colleen, welcome to the show.

[00:01:44] Coline: Thank you. Hi, Alex. Thanks for having me.

[00:01:46] Alexander: Yeah, very good. Maybe, introduce yourself to the listener and what you’re doing and what your how you came to this position at work.

[00:01:55] Coline: Yeah, sure. So again, thanks for having me on your podcast. And hi, everyone listening. So my name is Colleen Zeballos. I’m Swiss and French nationality wise, I’m married and I don’t have kids. That’s for the little introduction. Professionally, I’m our strategy lead at Roche Informatics actually since about three years, two and a half years, precisely. And I believe we’ll talk more about my job and its impact in this interview.

But before I, if you allow me to elaborate a bit more on my background, because even though you probably saw me regularly, probably in the past two years in conferences about, pharma regulatory submissions and R validation, health authority. I actually don’t have a STA statistical or biometrics background or education.

I actually come from business strategy management. , I did those kind of studies before joining Roche, and I have in terms of experience finance, data analysis, data science, and actually also entrepreneurship experiences and , so I worked for GlaxoSmith Klein or insurance, and then I created my own company before joining ro and I’m pretty sure that’s precisely.

this, that Rosh was looking for when they hired me. , this is strange or eclectic, not so typical set of skills and experience. And I can say now with distance that I think I bring something complimentary to the very technical and clinical experts that we hire at Rosh that are amazing and doing an excellent work.

But I, compliment them. And I rely on them actually rely on their expertise and I try to enable them to be better at what they do. So this is in short, and maybe to finish my introduction and because I believe that to be an accomplished professional, you also need to take time for your, out of work passions.

So very quickly, what I like to do outside of work is singing. I do opera, singing style. I’ve been

[00:03:49] Alexander: Wow. That’s cool.

[00:03:51] Coline: Music observatories. Yeah. For many years. I do some public performances from time to time. I won’t share anything right now, but yeah, that’s what I like to do. And also I started kite surfing a couple of years ago, so that’s another of my hobbies. Yeah.

[00:04:05] Alexander: Very cool. So that’s a really interesting combination. So let’s talk a little bit about your role. When I saw. The role of predator lead, I was really intrigued because on one hand, of course, r in itself is a hot topic Yeah. For a couple of years. And it’s only has becoming more hot over the last years.

The progress is amazing with more and more companies that invest in that. If we think about people coming from university, Most of them are well trained and are, there’s a lot of new releases in terms of our packages. Just recently there was a new release for basically a publishing platform from our studio.

There’s a lot of things going on. And also when I think about the wonderful Wednesday webinar series, which was also data visualization, nearly all of the submissions are in r and so just that just speaks to, the shift that we had of the years, because when I, maybe five, six years ago was.

Everything involved. And there, there has been a really dramatic shift to the other. Part of the role that have found really interesting is because I think in itself is a super interesting topic that I’ve over couple of years now, I had earlier as a podcast guest, I had person that was writing about.

How, a lot of the strategy has developed more in the military space and how we can learn from that port business area to that earlier, the show notes and also here, just blank on the name, . These two things really intrigued. So tell me a little bit more about what is an R strategy.

[00:06:04] Coline: Yeah, I’m happy to talk about it.

This is my daily life right now, and I it’s definitely a new role, I think in at Roche and I’m sure in, in other pharmaceutical companies. It, so I’m, I wanna emphasize the fact that I’m, our strategy lead in informatics. So at Ro Informatic, It’s a slightly different than being in the business, what we call it.

Usually we say informatics versus business. It really it has two different functions. So I’m in informatics and on the business side, you have also, you have the art developers, the biostatisticians, the leads, study leads, and we also have an r enablement lead actually to make sure that developers are trained and are, and get, so that’s on the business side, what I do in the informatics side, and I think.

Was quite strategic, a strategic decision and move from Roche two years ago. And when they hired me is really that they wanted someone to, to support the use of open source programming languages and to drive a strategy, lead us, have a strategy for supporting the use of open source programming languages in the business.

So we have to build a cohesive approach for. Because we can’t have, a different, for example, validation approach in each department or for each. If I exaggerate for each study, that would not be scalable and so on. So I play a role both internally and externally. So internally what I do, so externally and externally from Roche, right?

So internally I lead and coordinate. , how informatic supports the use of open source programming languages in business functions. As I said, for example, I bring solutions to questions like, how do we validate our packages for our clinical trial data analysis and submissions and R because as you rightly said, we, you said a shift from SAS to r.

I would say we maybe widened the scope of the programming languages we use, because I think sass will not disappear from the picture, but Okay. Let’s I think it’s also symbolic to say that we are shifting to r but SASS is still in the picture just to say, so I answer solutions like that, or also how do we manage the versions and dependencies of our packages.

For example, we implemented the use of a package manager tool to do that. That’s a new challenge. Just we didn’t have that with sas. So also, can we also validate Python packages? Okay. So things like this, I, here I listen, I take those needs and I build teams. I find solutions also something, for example, something that I did a year and a half ago, I built capabilities internally to validate our packages in an automated, And so we that this way we actually.

Dependencies with external vendors that were doing this for us and we built, we brought the capabilities inside Rosh. And I think that’s really powerful because there’s no reason why we won’t, we shouldn’t be able to validate our code ourselves. Cause at the end, validation is about making sure your code is a good quality and that it does what it’s supposed to do, right?

Yeah. So we actually bring the. This capability and building software quality inside, which is I think super valuable. This implied that I created a team, we, I give the vision and make sure that the team is delivering a solution that fits our internal needs and satisfies our internal customers.

So there’s a whole operational side of things that I then also delegate. Once I build it. But that’s one thing. I also extend extended the validation process and tools to Python. So we not just did R we’re doing Python as well. And I also, I think this is more, this is less tangible, but I also connect the dots between different teams, with similar needs to avoid silos.

We need really to make sure we have a cohesive strategy. Also, I need to ensure that my key internal customers, first of all, and second of all, the decision makers. In this company, at least in, in RO and pharma, that they adhere to my strategy. Because without the support of other decision makers, I can’t make a decision of, implementing a validation solution or investing in R and Tyson.

No. So alignment is super important. So I spend a lot of my time communicating, removing doubts, misunderstandings, listening to new needs, challenges, and so on. So that’s internally. .

Okay. Yeah. Go on with the external.

Exactly. Maybe also internally, something we did is we challenged and disrupted the traditional way of validation.

That was traditionally done. So we brought innovative ideas and also changed the way we did validation. Externally, I represent the work that we do at conferences, for example, are in pharma, but I also did a couple of others. I also voluntarily support and work for the our consortium.

That most of you must know as an active member. And I do see another series of initiatives to promote the use of. Cross industry and I help new conferences see the light of day by using my network. For example, we’re putting right now a series of events around promoting the use of R in the insurance industry.

For example, something that hasn’t really been tackled. And I’m regularly also contacted by similar roles that I have. Were actually my business counterparts in different companies. So we also talk about our different challenges and share.

Cool. What I really loved about your description of your role internally is that it really has all these different aspects of strategy.

Yeah. First really understanding the problem. Yeah. Like you just mentioned with a validation topic. Yeah. It’s not just, okay, we have done this validation this way, so we’ll do it the same way as. No understanding. Okay, what is really validation? What is optimally re required? And by having a different platform or maybe.

Just, 20 years, 30 years of experience with these kind thing is there’s something we can do differently. So just really understanding is the problem. And then kinda understanding also relationships. What kind bigger picture the problems or see kind what will happen takes all these kind of things into account and then, but not stop there, but also develop a technical plan, how to move it.

Challenging the status quo as well. I would add. You need the mindsets. You need people also internally to that have this mindset. Not everybody has it, so sometimes it’s more of a challenge than with other people. But it’s, we, you can make it. Yeah.

[00:12:28] Alexander: How often do you run in the sentence like we’ve always done? Yeah. ,

[00:12:32] Coline: A lot.

[00:12:33] Alexander: Yeah. For sure. , I can completely see so can you give an example for Hughes? Okay. Completely changed the status quo.

[00:12:41] Coline: So if example, with the validation, because we had a way of validating software in inside Rosh.

We had, the CSV validation way of doing, which was super, clearly defined with. And we were using a certain tool set of tool chain of. Tools yeah, sorry. A tool chain of tools or instead of tools to validate. And we, and it was super squared and there was not a lot of innovation and possibility to, and nobody actually, I think, ever.

Question the way we were doing validation because, validation is seen as this thing we have to do on top and we we just do it. Or we delegate to a team of validation lead and we just, the people, the experts don’t really care about validation. It’s actually really annoying.

We spoke a, with a lot of validation leads. We explained our perspective that. We need to integrate validation into the development process of an R package. Yeah. Because the r our language allows you to have embedded in the code your requirements, your tests, everything, all the information about the package is in the code.

So you don’t have to, have your requirements in Jira, for example, and extract those and then pull them into another. Everything is in the code. Yeah. You just you simply in between codes simply need to extract that, make sure that you have a test for each requirement without going into too much details, and then you produce the documentation that talks about what your package does and how we make sure that it does what it’s supposed to do.

All of this can be automated and is automated today, but we had to. Speak to a lot of people for that to happen is that,

[00:14:16] Alexander: I really love this example because it’s, as you said, every company has a procedures and it has always been done that way. And double programming whatsoever. Yeah. And thinking, ah, we, we can meet the same.

Goal, the same underlying goal in different steps and understanding part, what do we really do by these processes? What problem do we really solve? I think that is, is really part of strategy, thinking things completely new.

[00:14:47] Coline: Yeah. Yeah. And also because it’s allow me to add. If you create code of good quality, if you’re writing, you’re developer, you’re writing your package and adopt best practices for co-development you add your requirements, your test, you developed a good syntax and so on, then validation is easy because you have everything you need in your package.

And actually by, by, by creating this tool and by automating validation, we actually embedded validation into development, but also we naturally increase the quality of our code very naturally as well. So it’s a win-win.

[00:15:22] Alexander: Yeah. And that also means that shifting from soft to r it’s not just replicating what you have done with the same processes and everyth.

Just with a different language. It’s a mindset shift. There’s a lot of you opportunities. We’ve talked in the data visualization quite a lot about, that you shouldn’t do the same thing. It’s couldn’t be just about, oh, we only, we produce TFOs. No preproduced, TFOs that are, no, yeah, that’s not how it’s supposed to be.

It’s. Just one way to display an, inside number that just one way. Yeah. And in many ways, not an optimal way. It’s just, because we have been doing this for 20 years and we are used to. Sending big files and trucks was files to, to the fda. And now we do that and yeah, upload that data.

In the end, it’s still pretty much the same. All of our internal processes for all of these other things, we are not to a four size table. So that’s lot of other opportunity. That’s so cool. Yeah I love the validation example because it shows so clearly that it’s not so switching software. It’s having very different mindset about it.

[00:16:53] Coline: Yeah, no, exactly. And again I want to say that we’re not really, so I want to say, I think it’s important for, I’m representing what we do at Ross here. It’s true that we are shifting to the use of open source programming languages, but we’re not just erasing SaaS from the picture.

Actually, What we want to do. And I think because we think obviously we think it’s the right thing is we want to be able to say to our developers or our biostatisticians, use the language of your choice, depending on the situation, depending on the type of analysis you need to do, or, machine learning or whatever you need to do, depending also on your competencies and your capabilities.

So we want to give them this option and. and that’s something that’s that we’re

[00:17:37] Alexander: working on. Yeah, that’s the other thing, the people side of it. Yeah. You can’t just kinda switch from one day or one year to another kind of all those things because you need to take the people with you.

Absolutely. Speaking now about what has happened, let’s a little bit and speak about what might happen in the future, because that is one of the biggest interest things. Yeah. So if you just kinda. From what you have seen been doing and the last couple of years, what do you think will be possible in the future?

That is not possible now.

[00:18:16] Coline: I think we want to be, we’re doing all of this to be more efficient in our, in this space, in the space I work in the clinical trial. . different phases, right? We’re not just changing language or changing our ways of working just to change, right? So it’s at the end to be more efficient and to deliver drugs more efficiently and so on.

So from collecting the data to cleaning it, to transforming it into the c diff formats as dtm atom to performing statistical analysis, and then to submitting. We are, adopting this open source mindsets so that we can do this more efficiently. For example, we are able to replace the many thousands of tgs or the tables listings and graphs. Those outputs that are created by our biostatisticians and data analysts team. We generate, we can actually. View them. Use app-based Yeah. Views of our data. Which is much more efficient. Yeah. This shift opens a lot of doors in terms of also code reuse. . If our colleague in Novartis or in Pfizer or in another company has created a package to solve the same problem that we have, or to, improve one part of the process, we should be able to reuse it.

 And instead of creating it again. So that’s a dimension that open source programming language adoption enables actually. So again, to be more efficient

[00:19:36] Alexander: But that means also much more transparent. Isn’t. Yeah. As well. Yeah. So should, because what’s happening now is very often these, the code states within the companies just maybe share with the FDA that you don’t really see going on outside.

But when you know these kind of things are more public, you can see Exactly. Okay. For example, for imputing missing values or. Calculating certain things that are a little bit more complex, you can directly see, okay, this is exactly how they did it, so let’s do it exactly the same way so that it’s really consistent so that the apple from brush is the same apple from Pfizer’s, the same apple from gsk.

[00:20:22] Coline: So I think you don’t have to do exactly the same. And I think no company will do exactly, will satisfy itself with doing exactly what its neighbor is doing. You and those big actors in the pharma industry, you always want to add your piece of piece of, a bit of soft.

[00:20:36] Alexander: Yeah. But for if it’s about defining a certain endpoint so if you think treatment response or do they find it here?

You wanna have comparable data across different studies. Yeah. And so then of course these kind of things make a lot of change. Of course. In terms of, certain things will always be different. Study will say, oh, we want to use this primary endpoint, or different primary endpoint or later earlier, or different study population, all these kinda different things.

Yeah. You always optimized by your compounds of your compound. Certain kind of, understanding, okay, how did I do that? This will be, more, more transparent in the future.

[00:21:16] Coline: Yep. And it, I think it fosters innovation, don’t you think? Yeah.

[00:21:20] Alexander: Yes. The funny thing is standardization enables innovation doesn’t hinder that because then you could do the things that you don’t wanna innovate on fast.

As you have more time to think about the other things you than spend time to. Another demographic tables that you have done. 14 years. . Yeah. Yeah, exactly. What else do you think will be possible in the future? So

[00:21:47] Coline: I think I said already a number of things. So overall, So without going into the I won’t talk about the, so I’m not my self assess programmer or I’m, so I won’t do comparison like function by function comparison.

But at my level, at the strategic level, what I also see is that by opening up to open source programming languages and by, enabling. You to do submissions in SaaS and R and Python in whichever language you want. I think we can also, that’s better for each company because you, you more likely will put more molecules on the market, right?

So it also opens opportunities. Also, a side effect of that is that we work more with health authorities to enable this to happen. So there’s more collaboration, communication where maybe, whereas maybe in the past, it was just the status quo we were doing with SaaS health authorities were used to receiving SAS code and there was no innovation.

Maybe really? I exaggerate of course, but, so I think it also side effect is that it also allows more communication and innovation on the health authority side, and they also need to train give, give more capabilities to their workforce.

[00:22:54] Alexander: Yeah, instead of potentially, as you just said, you have these long list of tables that you wanna communicate and just printing them on paper on isn’t really effective.

So if you use internally an app, can you provide that app also to the regulators? Can they maybe, look into lots of different subgroups on their own way? Can. Instead of, working on getting all the tables validated, have more time for really good data visualizations, those interactives and explanatory data visualizations.

I think there’s a lot of opportunity to get away from the, let’s say, boring tasks, a repetitive task and free up time. The creative things, the innovative things that will be really cool. There’s one other thing that I think it could be very interesting is that the opportunity to connect the results together with their metadata.

Yeah. To let you direct, directly see. This number is a means on this endpoint at time point in this study, and so on. Yeah. And that will then later enable that you can use this beyond the regulatory submission, beyond the csr that you can use it also for publication, for HDA purposes, for clinical trials.

All kind of other areas. Yeah. So that it’s not okay. The accessibility to the results themselves completely changes and that, because what currently very happen is you have this PDFs output and terms. Someone says, oh, we need to have the nice graphic from it. There’s some poor guy. Through all the PDF number into and puts it into a PowerPoint and goes to somewhere else.

That’s another, like the same graphic, which kind different color. It just has the kind of looks into, oh, this is approximately, this is this number. A new cop, and there’s no traceability soever. And I think with this integration there, there will put potentially the opportunity to be much more both. Both forward and backward traceable.

Yeah. So that you have, in terms of forward traceability, panel, you can see this data was in all these different areas as well as if you. An output slide or whatever you can go back to that is where was those program. And that will help a lot for companies to understand what they actually know because they think many companies have so much knowledge that there’s nobody that oversees this.

Yeah, but this has been in many situations where I also spoke with colleagues about kinda, what lots of do later on in the lifecycle. This search for analysis that probably we have done, that we have done. That’s number here. So where the corresponding also Yeah. And that’s so ineffective.

[00:26:29] Coline: Yeah. Absolutely. It goes in hand. It goes hand in hand with the reusability of the data. So fi, findability of the data and traceability and reusability.

[00:26:39] Alexander: Yeah. Yeah, that’s right. Awesome. So that really lot points that we wanted to talk about. I love the strategic approach, that kind really understanding problem up to kinda implementing a solution and moving these things forward.

It means when we shift from , it’s not just, same stuff, different name, and there is a lot of opportunities for us in the future. That’s so cool. Yeah. And yeah, as you said, we need to take some people with us to, to make sure that Yeah, because there truly will be a very long transition.

[00:27:18] Coline: Yes. And there’s a lot of interest already. I can see it, with the audience that come to the events that the art consortium organizes at that, that I also bring together internally and externally. There are, there’s a lot of interest. We had about 600 Participants live in the last week we did an, our adoption series.

, event with three f d a speakers. And that was thanks to the work of my colleague Ning Lang from Ro and myself. And it was a huge 600 is a lot of participants for this type of event. So huge interests. People are watching the recording a lot, so you. I, so again, I want to address myself to all the people in different companies who work on this topic who help drive this this strategy because we need all of you.

And we also, I take this very seriously and I want to build common understanding on what we’re doing and bring people together around this this move to open.

[00:28:12] Alexander: Thanks so much Colleen. That was an outstanding last statement. Thanks for being on the show.

[00:28:18] Coline: Thank you alex.

Have you already requested access to the libraries that we have with all the free content around the effective statistician? If not. Head over to the effectivestatistician.com and look for this free library that all different nice things about it. Data visualizations of a in our recordings. Everything that I have for free. This show, was created in association with PSI. Thanks to Reine and her team at VVS will help with the show in the background. And thank you for listening reach your potential, lead great science and serve patients. Just be an effective statistician.

Never miss an episode!

Join thousends of your peers and subscribe to get our latest updates by email!

Get the shownotes of our podcast episodes plus tips and tricks to increase your impact at work to boost your career!

We won't send you spam. Unsubscribe at any time. Powered by ConvertKit