新藥發現是一個極其燒錢、低效的流程。候選藥在經歷了高成本、多年期臨床試驗流程之后，有90%最終都與市場無緣。90%的失敗率會拖累醫療產業的發展，同時也成為了大型公司為其天價藥辯解的托詞。事實證明，即便是在這個科技發展突飛猛進的年代，上述失敗率卻絲毫沒有松動的跡象。

讓我們想一下：自從美國前總統比爾·克林頓宣布完成首個人類基因組測繪之后已經過了近20年的時間。克林頓曾經說過，“這是科學與理性領域一個劃時代的勝利”，能夠帶來“大量新治療方案”，而且“可以通過攻擊疾病的基因來治療阿爾茨海默癥、帕金森綜合征、糖尿病和癌癥等疾病。”

雖然這些美好的目標依然只是正在進行的工作，但英國制藥巨頭葛蘭素史克的研發負責人哈爾·巴倫認為，該行業可能終于即將迎來具有實際意義而且卓有成效的基因革命。巴倫是一位心臟病學家，曾經在基因泰克和Alphabet的長生實驗室Calico擔任過領導職務，雖然他不大愿意做出這類籠統的聲明，但他在《財富》雜志上周于圣迭戈舉行的健康頭腦風暴會議上說，當前的形勢“給我一種不同的感受”。

他解釋說，作為一項基礎性工作，過去的基因組測序所提供的“構造性”洞見十分有限。科學家可以破解DNA，并發現堿基對鏈中的變異，但測序難以解釋隨之而來的生物學現象，也就是這些變異如何影響基因發揮作用的方式（以及后續會給攜帶這些基因的人群帶來什么影響）。

他對當前形勢十分樂觀的原因在于新出現的一些技術，即更加廉價的高通量DNA測序和深度學習。這些技術近期的同時興起也讓葛蘭素史克這類大型制藥公司能夠開展功能基因組學的研究。他解釋說，功能基因組學可以幫助建立基因變化與功能之間的聯系，解決近些年來出現的種種基因謎團（例如，為什么在出現特定變異的所有個體中，有30%患上了帕金森綜合征，但其他人卻沒有），并確立藥物靶標。巴倫說，有數據顯示，如果藥物靶標能夠得到基因驗證，那么新藥開發的成功率有可能會翻倍。

少數基因數據庫，從英國的Biobank到23andMe的自愿客戶基因池（葛蘭素史克去年夏天與這家基因檢測公司達成了一項協議），提升了驗證工作可行性。就深度學習而言，我們可以利用其能力分析人類生物構成方面“異常復雜”的多領域海量數據集。

這位研究負責人表示，功能基因組學還處于初期，葛蘭素史克的項目也不過是剛剛起步，但他計劃迅速向前推進。巴倫說：“我們希望成為業界的領軍者。”（財富中文網）

譯者：馮豐

審校：夏林

Drug discovery is a notoriously expensive and inefficient endeavor. Nine of 10 drug candidates that go through the cost-intensive, years-long clinical trial process never make it to market. That 90% failure rate—a drag on medical progress and a factor often cited by pharmaceutical companies to justify sky-high drug prices—has proven frustratingly sticky even in an era of breakneck scientific and technological advancements.

Consider: it’s been nearly two decades since President Bill Clinton announced the completion of the first-ever survey of the human genome, “an epic-making triumph of science and reason” that Clinton said would bring “immense, new power to heal” and “cure diseases like Alzheimer’s, Parkinson’s, diabetes, and cancer by attacking their genetic roots.”

Those lofty goals are still very much a work in progress, but Hal Barron, the head of R&D for the U.K.-based pharmaceutical giant GSK, thinks the industry may at last be on the cusp of a meaningful and productive genetic revolution. Though Barron, a cardiologist who previously held leadership roles at Genentech and Alphabet’s immortality lab Calico, hesitates to make such sweeping claims, he said at Fortune’s Brainstorm Health conference in San Diego last week that this moment “feels different.”

The genome sequencing efforts of the past, while foundational, offered limited “structural” insight he explained. Scientists could decode the DNA and identify mutations in a chain of base pairs, but the sequencing did little to explain the resulting biology—or how those mutations affect the way genes function (and in turn, impact the humans who hold them).

He’s optimistic about the current moment because of a handful of technologies—namely cheaper, high-throughput DNA sequencing and deep learning—have recently and simultaneously made the study of functional genomics feasible for big pharma companies like GSK. Functional genomics, he explained, can help link the genetic variant to function, solve the genetic mysteries that have piled up in recent years (why 30% of individuals with a certain mutation get Parkinson’s Disease, but the other 70% does not, for example) and pinpoint drug targets. Barron says data shows drug development is twice as likely to succeed when the drug target is genetically validated.

A handful of genetic databases, from the UK Biobank to 23andMe’s consenting customer pool (GSK did a deal with the genetic testing company last summer), have made the validation work more possible. As has deep learning, with its power to analyze the massive, multidimensional, “wildly complicated” data sets that make up human biology said Barron.

The research chief said it’s early days for functional genomics and that GSK’s program is in its infancy, but he plans to move quickly. Said Barron: “We’re hoping to be the leaders.”