Lost in Latent Space

Gene Drives, FAQ

What is a gene drive, in one sentence?

A genetic construct that increases its own probability of being inherited above the 50% Mendelian default, so that it spreads through a sexually reproducing population even when it confers no fitness benefit — and often despite a fitness cost.

Why isn’t standard Mendelian inheritance enough?

Because standard inheritance is symmetric: a parent passes each allele to 50% of offspring. If you want to replace a wild allele with an engineered one across an entire population, 50% won’t get you there without selection pressure. With strong negative selection (e.g. a sterility gene) the engineered allele drops out fast. A drive cheats by making the engineered allele the one that gets passed on, regardless of fitness — at least in principle.

How does a CRISPR homing drive actually work?

A homing drive is a CRISPR construct that encodes:

  1. Cas9
  2. A guide RNA targeting the wild-type version of the very locus where the construct sits.

When a heterozygous animal (one drive allele, one wild-type allele) is formed, the drive expresses Cas9+gRNA in the germline. Cas9 cuts the wild-type chromosome at the target. The cell repairs the break by homology-directed repair, using the drive-bearing chromosome as the template. The result: both chromosomes now carry the drive. The animal is effectively homozygous, and 100% of its offspring inherit the drive instead of the expected 50%.

That self-copying — the cell rewriting one chromosome to look like the other — is the entire mechanism. Everything else is engineering.

So why isn’t every wild mosquito already CRISPR’d?

Two reasons, both serious.

First, resistance. NHEJ (the error-prone repair pathway) competes with HDR. When NHEJ wins, the cut is sealed with an indel that destroys the gRNA target site — producing a mutant that is now permanently immune to the drive. In population terms: each generation produces resistance alleles, those alleles are positively selected (because the drive often imposes a fitness cost), and the drive fails. Early lab drives all collapsed this way.

Second, regulatory and political reality. No drive has been released into the wild. Even highly engineered “self-limiting” drives are contained to lab cages or remote islands. A drive in Anopheles gambiae that suppresses female fertility (the most-studied candidate, targeting malaria vectors) cleared lab cages but has not been approved for field release anywhere.

What’s a “self-limiting” drive?

A drive engineered so it cannot spread indefinitely. Examples:

Self-limiting drives trade reach for safety. None has been field-tested at scale.

Couldn’t a drive cause an ecological catastrophe?

Possibly. A drive that suppresses a population could collapse that species; a drive that escapes the target species (via horizontal gene transfer or interbreeding with a sister species) could spread further than intended. The most serious published concerns are not about the direct effect of the engineered phenotype but about the uncontrollability of a self-sustaining drive once released. There is no recall.

This is why the field has shifted hard toward self-limiting designs even though they are less effective.

Is there a positive use case people agree on?

Malaria. Anopheles gambiae is the primary vector across sub-Saharan Africa, malaria kills hundreds of thousands of children annually, and A. gambiae has no positive ecological role analogous to a pollinator. A drive that suppresses female fertility in A. gambiae would, on paper, solve a problem of staggering humanitarian weight. The technical readiness is plausibly there. The governance readiness is not. That asymmetry — a working tool that nobody is willing to use — is the current state of the field.

The original Burt 2003 paper that proposed homing drives, then the 2018 Kyrou et al. A. gambiae cage-trial paper, then anything Kevin Esvelt has written on daisy drives and the ethics of release.