TheBioMLClinic

TheBioMLClinic

TheBioMLClinic

ML for AAV Experimentalists Series

ML for AAV Experimentalists Series

ML for AAV Experimentalists Series


A series of six posts written for bench scientists and PIs who want to work with ML without outsourcing their scientific judgment to it.

Covers the core concepts, the experimental workflow, the main AAV-ML applications, the claims you will encounter in the field, and how to evaluate collaborations. No jargon.


A series of six posts written for bench scientists and PIs who want to work with ML without outsourcing their scientific judgment to it.

Covers the core concepts, the experimental workflow, the main AAV-ML applications, the claims you will encounter in the field, and how to evaluate collaborations. No jargon.


A series of six posts written for bench scientists and PIs who want to work with ML without outsourcing their scientific judgment to it.

Covers the core concepts, the experimental workflow, the main AAV-ML applications, the claims you will encounter in the field, and how to evaluate collaborations. No jargon.

Post 1: ML That Proposes New Sequences

Generative AI does not search your library, it proposes sequences you never built. This post explains how generative models work, when to use them with or without your own data, and what questions to ask before trusting what they produce. All in capsid engineering context.

Read the Post

Post 2: ML That Scores Your Variants

When you can only test a fraction of your candidates, how you prioritize determines what you find. This post covers how predictive models filter failures and rank variants, what they can and cannot reliably predict, and why the training context is always the right question to ask first.

Read the Post

Post 3: How ML Fits Into AAV Experimental Workflows

Knowing what ML tools do is not the same as knowing when to use them. This post maps generative and predictive models onto the real decision points in an AAV campaign: before synthesis, before screening, and after your first round of hits, so you know where they add value and where they do not.

Read the Post

Post 4: Where ML Is Being Applied in AAV Engineering + What to Expect

ML in AAV has expanded well beyond capsid design. This post surveys 11 application areas, from production fitness and tropism to manufacturing, vector genome quality, and agentic AI, with an honest maturity assessment for each so you can tell what is ready to use and what is still more concept than practice.

Read the Post

Post 5: How to Assess ML Claims in AAV Without Being an ML Expert

A 90% accuracy sounds rigorous. It may not be. This post gives you three diagnostic questions and five evaluative habits that cut through vendor pitches, conference abstracts, and collaborator proposals, so you can tell when the evidence is strong enough to act on and when it is not.

Read the Post

Post 6: Working With ML Teams, What AAV Scientists Actually Need to Know

Most AAV-ML collaborations fail not because of bad science but because the scaffold between the two sides was never built. This post is the reference guide for that scaffold: what to document before the data transfer, how to communicate experimental constraints as design specifications, how to push back on model outputs without derailing the relationship, and how to recognize a collaboration that is set up to fail before you have spent six months finding out.

Read the Post

We hope this gives you a clearer starting point.

We hope this gives you a clearer starting point.

— TheBioMLClinic

— TheBioMLClinic