Two years ago, as a Harvard undergraduate, I started a biotech company with a fellow student to develop an experimental drug for pancreatic cancer.

As biology majors, we had read a paper by a Harvard Medical School professor about an intriguing experiment and hoped we had found a possible solution to one of the world’s most lethal diseases. Rather than try to kill pancreatic tumors directly, we would block a cellular system called Hippo-YAP, which shields cancer from existing cancer-killing drugs. According to the paper, drugs that had been ineffective successfully destroyed pancreatic tumors in mice after Hippo-YAP was inactivated.

Could it work in people?

We partnered with the professor, launched Nivien Therapeutics and filed our first patent on Christmas Eve in 2016. We used my co-founder Nikita Shah’s Harvard dorm as our business address. A leading journal published the professor’s research. Scientists at the University of California at San Diego released complementary results demonstrating that our strategy might not only enhance existing drugs but also act as a novel immunotherapy, helping the body’s own immune system fight cancer.

In January 2017, we funded our first experiment with money we’d saved from work the previous summer. Soon, we won a grant from a business plan competition, raised $600,000 from venture capitalists and hired a scientific team. I took a leave from my senior year to work on Nivien full time, moving to San Francisco where we’d set up a new lab. Over the following months, working 90-hour weeks straight through the holidays, we created the first molecules to potently and selectively inhibit Hippo-YAP. With partners on three continents, we launched a barrage of experiments to test our hypothesis in preclinical trials.

It was an exhilarating time, with lots of positive press coverage — first in Harvard PR, then in the Daily Beast, Forbes and my hometown newspapers. I began receiving letters from patients and their friends and family.

“Recently, a dear 45-year old male friend was diagnosed with stage four pancreatic cancer,” wrote one woman. “He did have a successful Whipple surgery . . . but there is other involvement.”

Our chief medical adviser was a surgical oncologist at Massachusetts General Hospital with a PhD in pancreatic cancer. Every week, he does Whipple surgeries, which attempt to remove tumors from the head of the pancreas. He had taught me that just two words — “other involvement” — meant the writer’s friend would not make it to 50.

The writer asked whether we were offering studies that her friend could join. “He is looking to start treatment ASAP. . . . We who love him are desperate to see he gets a fighting chance at more years. . . . With the wonderful write-ups you are getting I’m hoping you are getting fully funded.”

Another note came directly from a patient:

“I am currently going through chemo for pancreatic cancer. . . . The treatment is working but it’s having severe side effects on my body. I was very intrigued by the research conducted by you guys . . . I was wondering if I can be any help to the cause.”

A third came from a tech executive two years into his own treatment: “I’d like to chat with you about the realities of your potential therapy, and even what opportunities may exist for a preclinical trial evaluation.”

Preclinical means before human studies, but it’s easy to be misled by how the media, and often biotech companies themselves, describe emerging science. All new discoveries take years to become new medicines — if they work at all — but even in scientific papers, it can sound like a cure is right around the corner.

The reality: Our potential therapy was another Hail Mary against a disease that has defeated doctors, patients and companies for centuries. While the lethality of most cancers is declining, the five-year survival rate in pancreatic cancer hovers at about 5 percent and the incidence of pancreatic cancer is rising.

Rereading an article about Nivien, I saw how easily a person with cancer could misinterpret the headline — and how I, like too many entrepreneurs, had not provided enough caveats about the potential of our work. Our strategy showed promise in mice but animal studies rarely translate into effectiveness in humans.

Still, some novel approaches do work, so it’s easy to get excited.

Raising money, recruiting great people and building a movement around your mission demands optimism, especially in failure-riddled drug development. Many biotechnologies really are revolutionary.

There’s an allure and a danger to believing in something like this, giving it your all, and convincing others to do the same while knowing the chances of success are lean. No one wants to commit to a marginal endeavor.

But employees want transparency. Investors want a fair bet. And patients want honesty. People with cancer are not one-liners for PR machines or slides for pitch decks. The philosopher Immanuel Kant wrote: “Treat humanity, whether in your own person or in the person of any other, never merely as a means to an end, but always at the same time as an end.” Now, whenever I read the latest and greatest biotech PR, I’m furious when I see oversold potential.

I spent hours replying to each of the letters we received. How to express reality without savaging hope? I scoured for available trials of new treatments, of which there were always too few. My responses sometimes led to exchanges that were at once inspiring and devastating.

In the end, our preclinical trials failed last summer. Inactivating Hippo-YAP did reduce the proteins protecting cancer, but the magnitude of benefit proved much less than we had hoped. Our initially positive results in mice didn’t translate into data that would save anyone.

Before Nivien, I’d met several people with pancreatic cancer. The father of a childhood friend set a record for survival with a certain type of the disease, dying after several years while we were in middle school. Once I left Harvard to work on Nivien, I didn’t meet a patient in person for months.

Then, during a visit home in August, my dad fell off a dock while helping his friend tie up a boat. Dad climbed out, glasses gone, and went in to dry. My girlfriend and I dove for the glasses until it got too dark to see underwater. We started talking with the skipper, whom I hadn’t met before.

He was tanned and fit. He also had late-stage pancreatic cancer. He’d heard about Nivien and wanted to learn more. I told him we were in a tough spot, explaining our results. Not as tough as his, I thought. He took the news about our data well — much better than I had. We invited him to join us for a hot dinner. He said he needed to move his boat to the lagoon before full dark, but that he’d call us after.

He didn’t come over for dinner. For months, I didn’t know if he was still alive. Then we saw each other again at the end of last year: He’d joined a clinical trial testing two immunotherapies and seems better off than many.

Now, when I email newly announced trials to the people who’d written us, and they don’t reply, is it because they’re too busy, too sick or already found a study? It’s seldom because they got better. Is it because they’re already dead?

Two years after starting Nivien, I told our investors we were done, helped our scientists find new jobs and returned to Harvard to finish my degree in molecular and cellular biology. Back at college I received one last message, a single line in an email: “My Dad has pancreatic cancer. Is there any way I can invest in your company as a private individual?”

The writer’s dad, it turned out, was only a couple years older than my own father.

I didn’t expect this outcome when we started Nivien. We wanted to build the next Genentech. We also knew from the start that defeat was the likely outcome. Nineteen in 20 drugs fail — after first-in-human trials.

Still, I’m proud of our work; that we took the shot. We developed the first molecules against a novel therapeutic target, broke new ground in the chemistry and biology of a key biochemical system and did it faster than industry standards despite our initial inexperience.

We ran the go/no-go experiment and the results were negative. No go. In this field unfortunately, that’s how it goes.