## 🌟 Data Science in Biotech Research Eric J. Ma Moderna --- ## 📝 About me > "Sr. Principal Data Scientist, DSAI (Research) @ Moderna" - Sr. Principal Data Scientist: Data science team lead <!-- .element: class="fragment" --> - DSAI: Computer nerd <!-- .element: class="fragment" --> - Research: Science nerd <!-- .element: class="fragment" --> --- ## 📅 Today > Explore what it takes to build a successful data science team in a biotech/pharma research organization. --- ## 🔍 Scope - Data science team - Biotech/pharma - Research organization --- ## 🎨 Format ```python themes = ["mission", "problem classes", "value delivery", "challenges"] for theme in themes: discuss() ``` --- ## 🌀 Format (cont'd) ```python def discuss(): preface() small_group_discussion() large_group_discussion() ``` --- ## 💡 Exchange of ideas - Discussion over lecturing <!-- .element: class="fragment" --> - Pragmatism over ideology <!-- .element: class="fragment" --> - Clarity over sophistry <!-- .element: class="fragment" --> - Commonalities and uniqueness <!-- .element: class="fragment" --> --- ## 🤝 I am here to engage in two-way learning with everyone --- ## 📋 Agenda 1. Mission <!-- .element: class="fragment" --> 2. Problem classes <!-- .element: class="fragment" --> 3. Value delivery <!-- .element: class="fragment" --> 4. Challenges <!-- .element: class="fragment" --> --- ## 📒 Notes hackmd.io/<span style="color: #6495ED">@ericmjl</span>/<span style="color: #40E0D0">ds-biotech-org</span> --- <!-- .slide: data-auto-animate --> ## 🎯 Mission > Why does your team of data scientists exist? ---- <!-- .slide: data-auto-animate --> > Why does your team of data scientists exist? - Charter? <!-- .element: class="fragment" --> - North star? <!-- .element: class="fragment" --> --- <!-- .slide: data-auto-animate --> ## ❓ Problem classes > Biotech/pharma operates on the holy trinity of indication-target-molecule - Indication: disease - Target: hypothesized gene/protein/RNA/cell/tissue that, if perturbed, will reverse the course of disease - Molecule: the entity that we introduce to perturb the target. --- <!-- .slide: data-auto-animate --> > What problems does your team solve? ---- <!-- .slide: data-auto-animate --> > What problems does your team solve? - In scope? - Out of scope? ---- ## 💊 Therapeutic hypothesis > {{ molecule }} will hit {{ target }} to reverse the course of {{ indication }} ---- > What methodologies? ---- <!-- .slide: data-auto-animate --> <span style="color: #6495ED">domain</span> $\times$ <span style="color: #40E0D0">methodology</span> ---- <!-- .slide: data-auto-animate --> <span style="color: #6495ED">domain</span> $\times$ <span style="color: #40E0D0">methodology</span> | | <span style="color: #6495ED">mRNA</span> | <span style="color: #6495ED">Proteins</span> | <span style="color: #6495ED">LNPs</span> | |---|------|----------|------| | <span style="color: #40E0D0">AI library design</span> | ✅ | ✅ | ✅ | | <span style="color: #40E0D0">Computer vision</span> | ✅ |❌ | ✅ | | <span style="color: #40E0D0">Probabilistic models + custom algorithms</span> | ✅ | ✅ | ✅ | --- <!-- .slide: data-auto-animate --> ## 🌐 Value delivery > What is the value of the problems that your team solves? ---- <!-- .slide: data-auto-animate --> > What is the value of the problems that your team solves? - Return on investment (ROI)? <!-- .element: class="fragment" --> - Priority level within the company? <!-- .element: class="fragment" --> What examples do you have, to the extent that it can be revealed? <!-- .element: class="fragment" --> ---- <!-- .slide: data-auto-animate --> > What frameworks can we use to estimate the ROI of our work? - Time saved by laboratory scientists vs. time needed to build solution <!-- .element: class="fragment" --> - Cost of experiments that would otherwise needed to be done <!-- .element: class="fragment" --> - Replacement of vendor software costs <!-- .element: class="fragment" --> What other questions would you ask? <!-- .element: class="fragment" --> --- ## 🚧 Challenges > What challenges will one face when building out a new data science team within a new biotech's research organization? ---- > Laboratory research processes are unstable, generating non i.i.d. data. Examples: - Condition scouting in antibody evolution. <!-- .element: class="fragment" --> - Cell type changes in protein efficacy assays. <!-- .element: class="fragment" --> - Change in vendor supplier for molecules. <!-- .element: class="fragment" --> How do we overcome this challenge? <!-- .element: class="fragment" --> ---- > Do we need a team of unicorns? Or do we need a team to cover the unicorn skillsets? - Scientific domain knowledge - Mathematical modelling skill - Software development skills - Communication skills - People skills Reference [essay on hiring](https://ericmjl.github.io/essays-on-data-science/people-skills/hiring/). ---- > Time is ticking; how do we demonstrate value early on? ---- > Research is a revenue-consuming organization that takes years to deliver a new product. How do we demonstrate that there is value delivered by the data science team in the interim? --- ## ⭐️ Thank you